Вы находитесь на странице: 1из 64

65

Congratulations, Andy Liu!


It is with great pleasure that we congratulate Andy Liu on yet another
award, con rming the high regard with which he is held in the international
mathematical community. CRUX with MAYHEM is very proud to have had
Andy Liu on its Editorial Board for many years.
The following can be seen on the World Wide Web at
http:

www.case.org awards canpoy.htm

The Council for Advancement and Support of Education USA


and
The Canadian Council for Advancement and Support of Education
1998-1999 CANADIAN PROFESSORS OF THE YEAR
OUTSTANDING UNIVERSITY PROFESSOR

66
Andy Liu joined the faculty of the University of Alberta Mathematical
Sciences department in 1980 and has been making a profound impact on the
institution and its students ever since. His passions include a commitment
to the study of mathematics and to developing innovative techniques that
allow him to share his knowledge with students of all ages. Students must
not settle into passive learning but must be challenged to participate in the
process," explains Liu.
Liu's success stems from his unique knack for presenting di cult concepts in a clear and logical manner. Former student William Willette said,
I have not been inspired to think and achieve by anyone more than Liu. He
is not the type of instructor who just gives answers all the time; he inspires
students to think." By providing practical examples of theoretical concepts,
Liu helps his students understand and learn rather than simply memorize.
Those of us working in the area of education of gifted and talented children have long considered Liu to be a resource par excellence in mathematics
education," said a colleague, Carolyn R. Yewchuk.
In the classroom, Liu uses his lively sense of humour to maintain students' interest. His pleasant demeanour creates a comfortable learning environment. In every class, I know the names of all the students by the time of
the midterm test," says Liu. He has even mastered the art of writing upside
down so students can follow his written explanations while meeting with him
at his desk.
Liu extends himself beyond the campus in a variety of ways. His onand o -campus lectures as well as the courses he has designed, re ect his
extraordinary talent. Liu is a strong supporter of mathematical competitions as a way to motivate and promote interest in mathematics. He has
participated in mathematical competitions on the local, provincial, national,
and international levels. He has provided training sessions for University of
Alberta undergraduate students, prepared training materials for a number of
national Mathematical Olympiad teams, and conducted training sessions for
the International Mathematical Olympiad IMO teams of Australia, Canada,
Hong Kong, Taiwan and the United States. In 1995, he received the IMO
Certi cate of Appreciation and in 1996 received the David Hilbert International Award for the promotion of mathematics worldwide. Liu explains,
I feel that the University is an integral part of the community, and its involvement must extend beyond the con nes of the campus. Also, learning
is a universal endeavour which transcends political boundaries, and the university is rst and foremost an international institution."

67

THE ACADEMY CORNER


No. 23
Bruce Shawyer

All communications about this column should be sent to Bruce


Shawyer, Department of Mathematics and Statistics, Memorial University
of Newfoundland, St. John's, Newfoundland, Canada. A1C 5S7
Here is the rst of four articles from the 1998 Canadian Undergraduate Mathematics Conference, held at the University of British Columbia in July 1998.

Abstracts R sum s
e e
Canadian Undergraduate Mathematics Conference
1998 | Part 1
Tracer Kinetic Modelling in Dynamic Positron Emission Tomography

Mark Andermann
McGill University

Dynamic positron emission tomography PET has often been used in the quantitative estimation of physiological parameters in vivo. For example, tracer kinetic
modelling of time-series data has been employed on a voxel by voxel basis using
various methods of non-linear regression analysis. However, due to the low signalto-noise ratio and large data sets present in PET studies, such methods are often
unstable and computationally intensive. In this lecture, mathematical theory underlying current and past methods which have attempted to increase the precision of
tracer kinetic modelling will be reviewed. Subsequently, a novel technique involving
principal components analysis, split-and-merge image segmentation using hypothesis testing with both heuristic and statistical likelihood tests, as well as some aspects
of spectral analysis will be introduced. Bene ts and disadvantages of this method
will then be discussed.
You Can Disprove That?

Shabnam Beheshti
McGill University

Prove or disprove the following: If two sets A and B of reals are homeomorphic, then they are both dense respectively meagre. How many times have you
been faced with a question like this on a nal? If you have experienced the wrath

68
of a professor who enjoys giving rigorous multiple choice" examinations, then perhaps the importance of studying and constructing counterexamples has been revealed
already. Every student of mathematics is faced with a choice: either conjuring up
a creative counterexample or waiting for divine intervention the latter sometimes
takes awhile. Many times, we implicitly assume intuitively sound but analytically
false properties of our system, forcing us to draw incorrect conclusions; the power
of counterexample is thus discovered. Such an error hindered progress towards the
proof of Fermat's Last Theorem. I will attempt to survey the search for a proof to
the theorem of Fermat and present a variation of Kummer's counterexample to the
purported proof due to Lam .
e
Trigonometry, Astronomy, and Computation:
The Historical Quest for an Elusive Constant

Abraham Buckingham
The King's University College

Ancient astronomy gave rise to some of the best historical mathematics, especially in the realm of trigonometry. Accurate trigonometric tables were critical
to good astronomy, and accurate trigonometric tables relied on an accurate value of
the sine of one degree for the majority of values. Unfortunately an exact value for
this sine is impossible to nd using ruler and compass methods alone. This problem was worked on by Claudius Ptolemy 100-175 AD and Jamshd al-K sh ?-1429

AD. Ptolemy's geometric method stood for 1200 years, but al-K sh was not satisa

ed and constructed an improved geometric estimation for the sine of one degree.
Still not satis ed, al-K sh went on to develop a xed point iteration scheme for the
a

calculation. I will describe and compare these methods within their context.
Invariants and shoelaces

Benoit Charbonneau
Universit du Qu bec a Montr al
e
e
e
In the Montreal phone directory, there are seven entries for Jacques Labelle.
If X 's name is Jacques Labelle, we need more information to identify him. But we
know for sure he is not Nicolas Bourbaki although this is a subtle question.
One eternal question of mathematics is Are two objects in fact the same object?
Up to what we call isomorphisms, some properties of objects don't change. We
call them invariants. After this talk, you won't look at life the same way!
Invariants et lacets de chaussures

Benoit Charbonneau
Universit du Qu bec a Montr al
e
e
e

Dans le bottin t l phonique de Montr al, il y a sept entr es pour Jacques Laee
e
e
belle. Si le nom de X est Jacques Labelle, on a besoin de plus d'information pour
l'identi er. On est cependant sur qu'il n'est pas Nicolas Bourbaki quoique la question est subtile.
Une question eternelle des math matiques est Est-ce que deux objets sont

e
identiques ?
Certaines propri t s des objets ne changent pas a isomorphisme pr s, on les
ee

e
appelle des invariants. Suite a cet expos , vous ne verrez plus jamais la vie de la m^ me

e
e
fa on !
c

69
Ranking the Participants in a Round-Robin Tournament

Susan Marie Cooper


University of Regina

Various schemes for ranking and comparing the participants in a round-robin


tournament have been proposed. However, none of these are considered entirely
satisfactory. We will consider the Kendall-Wei method, a method of ranking tournaments using iterated strength vectors, which leads to the Perron vector i.e. vector
of relative strengths of the corresponding dominance matrix of the tournament. We
will investigate such questions as: how easily can the iterated strength vectors be
calculated?" and can we determine the relative ordering of the strengths of the participants without calculating the Perron vector?"
What is the 57; 5-cage?

Jennifer de Kleine
University of Northern British Columbia

An n; 5-cage is an n-regular girth-5 graph of smallest order. I will discuss


what this means, and the open question of whether there exists a 57; 5-cage with
572 + 1 vertices. Time permitting, I will discuss the connection between estimating
the permanent of 0; 1 matrices and a possible approach to using a computer to
search for a 572 + 1 vertex 57; 5-cage.
Les nombres de Stirling

Caroline Desjardins
Universit du Qu bec a Montr al
e
e
e

Jacob Stirling, math maticien anglais du 18e si cle a publi un article int ressant
e
e
e
e
Methodus di erentialis". L'article contient deux triangles de nombres entiers apparent s au triangle de Pascal dont les diverses composantes permettent d'obtenir
e
des r sultats int ressants en analyse, en combinatoire et en probabilit s. Ce sont
e
e
e
ces nombres que l'on appelle les nombres de Stirling. On peut obtenir ces nombres
en trouvant les matrices de changement de base de deux base de R x ou bien,
en utilisant les relations de r currences que ces nombres suivent. Les liens reliant
e
ces nombres aux nombres de partition en k-classes et de fonction surjectives d'un
ensemble a n el ments vers un ensemble a k el ments, sont quelquesexemples d'ap e
e
plication des nombres de Stirling.
Applications of Group Theory to Chemistry

Norman Dreger
University of British Columbia

The goal of this paper is to present a number of chemical applications of group


theory. Some key terms will be de ned and the concepts of symmetry elements and
symmetry groups introduced. The nomenclature of symmetry elements with regards
to chemistry will then be discussed. Finally some applications of symmetry to experimental chemistry will be examined. Only a rudimentary knowledge of group theory
is prerequisite.

70

THE OLYMPIAD CORNER


No. 196
R.E. Woodrow

All communications about this column should be sent to Professor R.E.


Woodrow, Department of Mathematics and Statistics, University of Calgary,
Calgary, Alberta, Canada. T2N 1N4.
We lead o this issue with the problems of the 19th Austrian-Polish
Mathematics Competitions, written in Poland, June 26 28, 1996. My thanks
go to Ravi Vakil, Canadian Team Leader to the IMO at Mumbai as well as to
regular supporters Marcin E. Kuczma, Warszawa, Poland and Walther Janous,
Ursulinengymnasium, Innsbruck, Austria for supplying copies of the contest
material.
19th AUSTRIAN-POLISH MATHEMATICS

COMPETITION 1996

Problems of the Individual Context


June 26 27, 1996 Time: 4.5 hours

1. Let k  1 be an integer. Show that there are exactly 3k,

positive
integers n with the following properties:
a The decimal representation of n consists of exactly k digits.
b All digits of n are odd.
c The number n is divisible by 5.
d The number m = n has k odd decimal digits.
5
2. A convex hexagon ABCDEF satis es the following conditions:
a The opposite sides are parallel; that is, AB kDE , BC kEF , CDkFA.
b The distances between the opposite sides are equal; that is,
dAB; DE = dBC; EF  = dCD; FA, where dg;h denotes the
distance between lines g and h.
c FAB and CDE are right angles.
Show that diagonals BE and CF intersect at an angle of 45 .
3. The polynomials Pnx are de ned recursively by P0x = 0,
P1x = x and
Pnx = xPn,1x + 1 , xPn,2x for n  2 .
For every natural number n  1 nd all real numbers x satisfying the equation Pn x = 0.
1

71

4. The real numbers x, y, z, t satisfy the equalities x + y + z + t = 0


and x2 + y 2 + z 2 + t2 = 1. Prove that ,1  xy + yz + zt + tx  0.
5. A convex polyhedron P and a sphere S are situated in space in
such a manner that S intercepts on each edge AB of P a segment XY with
AX = XY = Y B = 1 AB. Prove that there exists a sphere T tangent to
3
all edges of P .
6. Natural numbers k, n are given such that 1 k n. Solve the
system of n equations
x3  x2 + x2+1 +    + x2+k,1 = x2,1 for 1  i  n
i
i
i
i
i
with n real unknowns x1 , x2 , : : : , xn . Note: x0 = xn , xn+1 = x1 ,
xn+2 = x2 , and so on.
Problems of the Team Contest Poland
June 28, 1996 Time: 4 hours

7. Show that there do not exist non-negative integers k and m such


that k! + 48 = 48k + 1m .
8. Show that there is no polynomial P x of degree 998 with real
coe cients satisfying for all real numbers x the equation
P x2 , 1 = P x2 + 1 .

9. We are given a collection of rectangular bricks, no one of which is


a cube. The edge lengths are integers. For every triple of positive integers
a; b; c, not all equal, there is a su cient supply of a  b  c bricks. Suppose
that the bricks are completely tiling a cubic 10  10  10 box.
a Assume that at least 100 bricks have been used. Prove that there exist
at least two bricks situated in parallel, in the sense that if AB is an edge of
one of them and A0 B 0 is an edge of one of the other, and if AB kA0 B 0 , then
AB = A0 B0.
b Prove the same statement for a number less than 100 of bricks used.
The smaller number, the better the solution.
Next we move to a country whose contest materials have not been very
often available in CRUX with MAYHEM with the problems of the 3rd Turkish
Mathematical Olympiad, Second Round, written December 8 9, 1995. My
thanks go to Ravi Vakil, Canadian Team Leader to the IMO at Mumbai for
collecting the problems.

72

3rd TURKISH MATHEMATICAL OLYMPIAD


Second Round First Day

December 8, 1995 Time: 4.5 hours

1. Let a1; a2; : : : ; ak and m1; m2; : : : ; mk be integers with 2  m1


and 2mi  mi+1 for 1  i  k , 1. Show that there are in nitely many
integers x which do not satisfy any of the congruences
x ai mod m1 , x a2 mod m2 , : : : , x ak mod mk  .
2. For an acute triangle ABC , k1, k2, k3 are the circles with diameters BC , CA , AB , respectively. If K is the radical centre of these
circles, AK k1 = fDg, BK k2 = fE g, CK k3 = fF g and
AreaABC  = u, AreaDBC  = x, AreaECA = y , and AreaFAB  = z ,
show that u2 = x2 + y 2 + z 2 .
3. Let N denote the set of positive integers. Let A be a real number
and fan g1 be a sequence of real numbers such that a1 = 1 and
n=1
1

an+1  A for all n 2 N .


an

a Show that there is a unique non-decreasing surjective function k : N ! N


k
such that 1 Aann  A for all n 2 N.
b If k takes every value at most m times, show that there exists a real
number C 1 such that C n  Aan for all n 2 N.
 

Second Round Second Day

December 9, 1995 Time: 4.5 hours

4. In a triangle ABC with jABj 6= jAC j, the internal and external


bisectors of the angle A intersect the line BC at D and E , respectively. If the
feet of the perpendiculars from a point F on the circle with diameter DE to
the lines BC , CA, AB are K , L, M , respectively, show that jKLj = jKM j.
5. Let tA denote the sum of elements of A for a non-empty subset
A of integers, and de ne t  = 0. Find a subset X of the set of positive
integers such that for every integer k there is a unique ordered pair of subsets
Ak; Bk of X with Ak Bk = and tAk , tBk  = k.
6. Let N denote the set of positive integers. Find all surjective
functions f : N ! N satisfying the condition
for all m, n 2 N.

m j n == f m j f n

73
Along with the Turkish Olympiad we have the questions of the Turkish
Team Selection Examination for the 37th IMO, written March 23 24, 1996.
Thanks again go to Ravi Vakil, Canadian Team Leader to the IMO at Mumbai
for forwarding these to me.

TURKISH TEAM SELECTION EXAMINATION FOR


THE 37th IMO
First Day | March 23, 1996
Time: 4.5 hours

1. Let Qn 1 + nx n = 1 + a xk + a xk +    + amxkm where


a , a , : : : , am are non-zero and k k    km . Find a .
^
2. In a parallelogram ABCD with mA 90 , the circle with diam1996
=1

1996

eter AC intersects the lines CB and CD at E and F besides C , and the


tangent to this circle at A intersects the line BD at P . Show that the points
P , F , E are collinear.
3. Given real numbers 0 = x1 x2    x2n x2n+1 = 1 with
xi+1 , xi  h for 1  i  2n, show that

1,h
2

n
X

x2ix2i+1 , x2i,1  1 + h .
2
i=1

Second Day | March 24, 1996


Time: 4.5 hours

4. In a convex quadrilateral ABCD, AreaABC  = AreaADC  and


AC BD = fEg, and the parallels from E to the line segments AD ,
DC , CB , BA intersect AB , BC , CD , DA at the points K , L,
M , N , respectively. Compute the ratio
Area KLMN 
.
Area ABCD
5.

Find the maximum number of pairwise disjoint sets of the form

Sa;b = fn2 + an + b : n 2 Zg with a, b 2 Z.

6. For which ordered pairs of positive real numbers a; b is zero the
value of the limit of every sequence fxn g satisfying the condition
nlim axn+1 , bxn  = 0 ?
!1

74
To round out the contests for your puzzling pleasure we give the two
papers of the Australian Mathematical Olympiad 1996. My thanks go to
Ravi Vakil, Canadian Team Leader of the IMO at Mumbai, once again, for
providing me with the contest materials.

AUSTRALIAN MATHEMATICAL OLYMPIAD 1996


Paper 1

February 6, 1996 Time: 4 hours

1. Let ABCDE be a convex pentagon such that BC = CD = DE


and each diagonal of the pentagon is parallel to one of its sides. Prove that
all the angles in the pentagon are equal, and that all sides are equal.
2. Let px be a cubic polynomial with roots r1, r2, r3. Suppose that
p , 1  + p ,, 1  = 1000 . Find the value of 1 + 1 + 1 .
2
2
p0

r1r2 r2r3 r3r1

3. A number of tubes are bundled together into a hexagonal form:


f f f f
f f f f f
f f f f f f
f f f f f f f
f f f f f f
f f f f f
f f f f

A number of tubes in the bundle can be 1, 7, 19, 37 as shown, 61, 91; : : : .
If this sequence is continued, it will be noticed that the total number of tubes
is often a number ending in 69. What is the 69th number in the sequence
which ends in 69?
4. For which positive integers n can we rearrange the sequence
1, 2, : : : , n to a1, a2, : : : , an in such a way that jak , kj = ja1 , 1j 6= 0
for k = 2, 3, : : : , n?

Paper 2

February 7, 1996 Time: 4 hours

5. Let a1, a2, : : : , an be real numbers and s a non-negative real


number such that
i a1  a2      an ;
ii a1 + a2 +    + an = 0;
iii ja1 j + ja2 j +    + jan j = s.
Prove that
an , a1  2s .
n

75

6. Let ABCD be a cyclic quadrilateral and let P and Q be points on


the sides AB and AD respectively such that AP = CD and AQ = BC . Let
M be the point of intersection of AC and PQ. Show that M is the mid-point
of PQ.
7. For each positive integer n, let n denote the sum of all positive
integers that divide n. Let k be a positive integer and n1 n2    be an
in nite sequence of positive integers with the property that ni , ni = k
for i = 1, 2, : : : . Prove that ni is a prime for i = 1, 2, : : : .
8. Let f be a function that is de ned for all integers and takes only the
values 0 and 1. Suppose f has the following properties:
i f n + 1996 = f n for all integers n;
ii f 1 + f 2 +    + f 1996 = 45.
Prove that there exists an integer t such that f n + t = 0 for all n for which
f n = 1 holds.
Now, an alternate and more general solution to problem 2 of the Dutch
Mathematical Olympiad, Second Round, 1993 than the one given in the Corner in the October 1998 number 1997: 197 , 1998: 330 .
2. Given a triangle ABC , A = 90 . D is the mid-point of BC , F
is the mid-point of AB , E the midpoint of AF and G the mid-point of FB .
AD intersects CE, CF and CG respectively in P , Q and R. Determine the
ratio P Q .
QR
Cq
q qD
qR

Pq Q
q
q
qB
Aq
E Fq G
Alternate Solution by Geo rey A. Kandall, Hamden, Connecticut, USA.
We rst establish the following:
Lemma.
PQ = CP  EF  CG .
QR CE FG CR

Proof.

PQ = CPQ = CPQ  CEF  CFG


QR
CQR
CEF CFG CQR
CP  CQ  EF  CF  CG = CP  EF  CG .
= CE  CF FG CQ  CR
CE FG CR

76
We now solve the problem, without using the hypothesis that A = 90 .
By the lemma

PQ = CP  EF  CG = CP  CG .
QR CE FG RC CE RC

By Menelaus' Theorem we have

CD  BA  EP = 1 , hence EP = 1 ,
DB AE PC
PC 4

CP = 4 ; 1
CE 5

CD  BA  GR = 1 , hence GR = 3 , CG = 7 . 2
DB AG CR
CR 4
CR 4
PQ = 4  7 = 7 .
Consequently
QR 5 4 5
This method can be used with di erent ratios CD : DB and AE : EF :
FG : GB.
After the February number was nalized we received a package of solutions from Michael Selby, University of Windsor, Windsor, Ontario. This
included solutions to problems 1 through 4 of the Croatian National Mathematics Competition 4th Class May 13, 1994 for which the problems were
given 1997: 454 and the solutions 1999: 12 . He also sent a solution to
a problem of the Additional Competition for the Olympiad of the Croatian
National Mathematical Competition, given 1997: 454 .
1. Find all ordered triples a; b;c of real numbers such that for every
three integers x, y , z the following identity holds:

jax + by + czj + jbx + cy + azj + jcx + ay + bzj = jxj + jyj + jzj .

Solution by Michael Selby, University of Windsor, Windsor, Ontario.


Set x = y = z = 1; we obtain ja + b + cj = 1
1
Set x = 1; y = z = 0 we obtain jaj + jbj + jcj = 1
2
Set x = 1; y = ,1, z = 0 we obtain ja , bj + jb , cj + jc , aj = 2
3
This system is symmetric. Without loss of generality we may assume
a  b  c.
Now 3 becomes 2a , c = 2 or a , c = 1. Substituting into 1 and
2 gives

j1 + b + 2cj = 1

4

77
and

j1 + cj + jbj + jcj = 1 .

5

Squaring 4 and expanding gives

1 + b + 2c2 + 2b + 2c = 1 .


Thus b + 2c = 0 or b + 2c = ,2.
If b + 2c = 0, then from 5
j1 + cj + 3jcj = 1 .
Since jcj  1, 1 + c  0, therefore 1 + c + 3jcj = 1 and c + 3jcj = 0. If
c  0, we have 4c = 0 and then c = 0. If c  0, ,2c = 0 giving c = 0.
Therefore b = ,2c = 0, a = 1 + c = 1, in this case.
In case b + 2c = ,2, substitution into 5 yields
j1 + cj + 2j1 + cj + jcj = 1 .
Since 1 + c  0, 31 + c + jcj = 1. If c  0, 3 + 4c = 1 and c = ,1 . This
2
is impossible.
If c  0, 3 + 3c , c = 1 giving c = ,1. Then b = 0 and a = 1 + c = 0.
Therefore we have the solution a = 0, b = 0, c = ,1, and these are the
solutions for a  b  c.
Hence there are six solutions

1; 0; 0 , ,1; 0; 0 , 0; 1; 0 , 0; ,1; 0 , 0; 0; 1 , 0; 0; ,1 .


Next we turn to solutions by the readers to problems of the 17th AustrianPolish Mathematics Competition given in the February 1998 number 1998: 4 .

17th AUSTRIAN POLISH MATHEMATICS


COMPETITION
Poland, June 29 July 1, 1994
1. The function f : R ! R satis es for all x 2 R the conditions

f x + 19  f x + 19 and f x + 94  f x + 94 .


Show that f x + 1 = f x + 1 for all x 2 R.

78
Solutions by Michel Bataille, Rouen, France; by Pierre Bornsztein,
Courdimanche, France; by Pavlos Maragoudakis, Pireas, Greece; and by
Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario. We give
the solution by Bataille.
Let x be an arbitrary real number. Applying the given conditions to
x , 19 and x , 94 respectively, we obtain
f x , 19  f x , 19 and f x , 94  f x , 94 .
Now an easy induction shows that for all n 2 N,
f x + 19n  f x + 19n , f x + 94n  f x + 94n ,
f x , 19n  f x , 19n , and f x , 94n  f x , 94n .
Since 1 = 5  19 , 94 and 1 = 18  94 , 89  19, we get:

and

f x + 1 = f x + 5  19 , 94  f x + 5  19 , 94
 f x + 5  19 , 94
= f x + 1 ,

f x + 1 = f x + 18  94 , 89  19  f x + 18  94 , 89  19
 f x + 18 + 94 , 89  19
= f x + 1 ,
so that f x + 1 = f x + 1, as required.

Comment: the same result can be obtained from the more general hypothesis: for all x 2 R, f x + a  f x + a and f x + b  f x + b
where a and b are positive relatively prime integers. Indeed, the preceding proof adapts easily as we can nd positive integers m, n, p, q such that
ma , nb = 1 and pb , qa = 1.
2. The sequence fang is de ned by the formulae

a0 = 1 and an+1 = 1 2an 2 for n  0 ,


2
+a
n

and the sequence fcn g is de ned by the formulae


c0 = 4 and cn+1 = c2 , 2cn + 2 for n  0 .
n
Prove that
an = 2c0c1 : : : cn,1 for all n  1 .

cn

Solutions by Michel Bataille, Rouen, France; by Pierre Bornsztein,


Courdimanche, France; by Murray S. Klamkin, University of Alberta, Edmonton, Alberta. We give the solution of Klamkin, which gives an indication
of both types of solutions received.

79
Letting xn = cn n 1, we have xn+1 = x2 where x0 = 3. Hence,
,
n
xn = x2n and cn = 32 + 1. Since c1 = 10 and a1 = 4 it now su ces to
0
5
2an
show that an = 2c c c:::cn, satis es the recurrence an+1 = 1+an for n  0.
n
,
Also since 32n + 132n , 1 = 32n , 1, it follows multiplying by 3 ,1 
3
1
that
2n
0 1

20
20

+1

2c0c1 : : : cn,1 = 3 , 1
cn
32n + 1

and by substitution and simpli cation, this satis es the recurrence relation
for an .
Comment: We can obtain another representation for an by letting it
equal tanh n , so that tanh n+1 = tanh 2 n , subject to 1 = tanh 0 . It
2
,
,

then follows that an = tanh 2n 0 = tanh 2n arctan h 1 = tanh 2n,1 ln 3 .
2
4. Let n  2 be a xed natural number and let P0 be a xed vertex
of the regular n + 1 gon. The remaining vertices are labelled P1 , P2 , : : : ,
Pn, in any order. To each side of the n + 1 gon assign a natural number as
follows: if the endpoints of the side are labelled Pi and Pj , then ji , j j is the
number assigned. Let S be the sum of all the n + 1 numbers thus assigned.
Obviously, S depends on the order in which the vertices have been labelled.
a What is the least value of S available for xed n?
b How many di erent labellings yield this minimum value of S ?
Solution by Pierre Bornsztein, Courdimanche, France.
P0 Pi1

Pn

Pi2

Pik

a Soit P0 Pn l'arc r liant P0 a Pn dans le sens des aiquilles d'une


e

montre, P0 Pn  l'arc r liant P0 a Pn dans le sens contraire.
e

Notons S , la somme des nombres assign ssur P0 Pn idem pour S + .
e
Par d nition,
e

S, = j0 , i1j + ji1 , i2j +    + jik,1 , ikj + jik , nj


 j0 , i1 + i1 , i2 +    + ik,1 , ik + ik , nj = n
avec egalit ssi 0  i1  i2     ik n.
e

80
De m^ me,
e

S+  n

avec egalit ssi les sommets sont class s dans l'ordre croissant de 1 a n, d'ou
e
e


on en d duit S = S , + S +  2n.
e

b Pour Pn x il y a i sommets entre P0 et Pn , le long de P0 Pn ou
e
in2 10, : : : , n , 1g. Il y a donc i nombres a choisir dans f1, : : : ,n , 1g, d'ou
f


, ,
choix.
i
Les nombres, une foix choisis, sont alors dispos s dans l'ordre croissant
e
de P1 a Pn : l'ordre est donc impos .

e
De m^ me sur P0 Pn  les nombres restants sont impos s ainsi que leur
e
e
ordre.

P , ,
Il y a donc n=01 n,1 = 2n,1 choix pour la disposition.
i
i
5. Solve the equation

1 x + yy + zz + x + x + y + z 3 = 1 , xyz


2

in integers.
Solutionsby Michel Bataille, Rouen, France; by Pierre Bornsztein, Courdimanche, France; and by Murray S. Klamkin, University of Alberta, Edmonton, Alberta. We give the write-up of Bataille, although all three solvers used
the same approach.
Let s = x + y + z and

P X  = X , xX , yX , z
= X 3 , sX 2 + xy + yz + zxX , xyz .
Then x + y y + z z + x = P s = sxy + yz + xz  , xyz and the given
equation may be written

sxy + yz + xz , xyz + 2s3 = 2 , 2xyz ,

or 2 + P ,s = 0.
As P ,s = ,2x + y + z 2y + z + x2z + x + y , the equation
nally becomes
2x + y + z2y + z + x2z + x + y  = 2 .
Either one of the three factors of the left-hand side is 2 and the other two
are 1, 1 or ,1, ,1 or one of the factors is ,2 and the other two are 1, ,1,
or ,1, 1.
The system
8
:

2x + y + z = 2
x + 2y + z = 1
x + y + 2z = 1

is equivalent to

x = 1, y = 0, z = 0.

81
The system
8

2x + y + z = 2
x + 2y + z = ,1
:
x + y + 2z = ,1

is equivalent to

x = 2 , y = ,1 , z = ,1 .

When one of the factors is ,2, the two corresponding systems lead to
4x + y + z = ,2, which is impossible for integral x, y , z .
Since x, y , z have symmetrical roles, there are six solutions altogether
for the triple x; y;z :
1; 0; 0 , 0; 1; 0 , 0; 0; 1 , 2; ,1; ,1 , ,1; 2; ,1 , ,1; ,1; 2 .

7.

Determine all two-digit in decimal notation natural numbers

n = ab10 = 10a + bb a  1 with the property that for every integer


x the di erence xa , x is divisible by n.

Solutions by Pierre Bornsztein, Courdimanche, France; and by Edward


T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario. We give Wang's
solution.
Clearly, n j xa , xb for all integers x if a = b. We show that besides
11, 22, : : : , 99 there are exactly three more such n's. These are: n = 15, 28,
and 48. We assume that a 6= b and start o by eliminating some impossible
values of n.
1 If a is even and b is odd, then setting x = 2 leads to n j 2a , 2b and
n j 2a +2b. Thus n j 2a+1, which is clearly impossible since the only possible
divisors of 2a+1 are powers of two while n 1 is odd.
2 If a is odd and b is even, then setting x = 2 again leads to the same
conclusion that n j 2a+1 . Hence n must be a power of two. Since a is odd,
the only possible values are n = 16 and 32. However, 16 6 j 2 , 26 and
32 6 j 23 , 22 , showing that there are no solutions in this case either.
3 If b = 0, then n is even and n j 2a , 1, which is clearly impossible.
Using 1, 2, and 3 we narrow the possible values of n down to the
following set of 32 integers:
f13 , 15 , 17 , 19 , 24 , 26 , 28 , 31 , 35 , 37 , 39 , 42 , 46 , 48 , 51 , 53; 57 ,
59 , 62 , 64 , 68 , 71 , 73 , 75 , 79 , 82 , 84 , 86 , 91 , 93 , 95; 97g .
Since n j xa , xb if and only if n j xb , xa we may assume that a b when
checking whether n satis es the given property. Note that
23 , 2 = 6 eliminates 13 and 31 ;
24 , 22 = 12 eliminates 24 and 42 ;
25 , 2 = 30 eliminates 51 but not 15 ;
25 , 23 = 24 eliminates 35 and 53 ;
26 , 22 = 60 eliminates 26 and 62 ;
26 , 24 = 48 eliminates 46 and 64 ;

82

27 , 2 = 126
27 , 23 = 120
27 , 25 = 96
28 , 22 = 252
28 , 24 = 240
28 , 26 = 192
29 , 2 = 510
29 , 23 = 504
29 , 25 = 480
29 , 27 = 384

eliminates 17 and 71 ;
eliminates 37 and 73 ;
eliminates 57 and 75 ;
eliminates 82 but not 28 ;
eliminates 84 but not 48 ;
eliminates 68 and 86 ;
eliminates 19 and 91;
eliminates 39 and 94 ;
eliminates 59 and 95 ;
eliminates 79 and 97 .
Therefore, the only possible values of n are: n = 15, 28 and 48. We now
show that they indeed satisfy the condition that n j xa , xb for all integers x.
a For n = 15, we show that x x5 mod 15. By Fermat's Little Theorem
Fthm, we have x3 x mod 3 and so x5 x3 x mod 3. Also,
x5 x mod 5. Hence x5 x mod 15 follows.
b For n = 28, we show that x2 x8 mod 28. Note that 28 = 22  7.
By Fthm, we have x7 x mod 7 and so x8 x2 mod 7. Further, we
claim that x8 x2 mod 4. This is obvious if x is even. On the other
hand, if x is odd, then x2 1 mod 4 implies x8 1 mod 4 and so
x8 x2 mod 4. Hence x8 x2 mod 28 follows.
c For n = 48, we show that x4 x8 mod 48. Note that 48 = 24  3.
By Fthm, we have x3 x mod 3 and so x4 x2 mod 3. Hence
x8 x4 mod 3. It remains to show that 16 j x8 , x4. This is clear if
x is even. If x is odd, then x = 2k + 1 for some integer k and thus

x8 , x4 = x4 x2 , 1x2 + 1  ,
= 2k + 14 ,4k2 + 4k 4k2 + 4k + 2
= 8kk + 1 ,2k2 + 2k + 1 2k + 14 ,
which is divisible by 16 since kk + 1 is even.
To summarize, n = 10a + b satis es n j xa , xb for all integers x if
and only if n = 11, 22, : : : , 99, 15, 28, 48.
Comment: This is one of the most intriguing problems that I have seen
lately. I will be really surprised if there is a much shorter solution!
8. Consider the functional equation f x;y = a f x;z + b f y;z
with real constants a, b. For every pair of real numbers a, b give the general
form of functions f : R2 ! R satisfying the given equation for all x, y ,
z 2 R.
Solution by Pierre Bornsztein, Courdimanche, France.
Soient a, b 2 R et pour tous x ,y ,z 2 R
f x;y = af x;z + bf y;z .

Alors :

83
Dans le cas ou x = y = z , f x; x = a + bf x;x donc a + b = 1 ou

f x; x = 0. Si a + b 6= 1, pour tout x 2 R, f x;x = 0 et donc pour z = y,
 donne
f x;y = af x;y + bf y;y = af x; y .
Donc soit a = 1 ou f x;y  = 0.
Dans le cas ou a = 1
f x;y = f x;z + bf y;z ,
observons qu'avec x = y , f x;x = 0 = f y;z 1 + b, et donc f 0 ou
b = ,1.
Maintenent si a = 1 et b = ,1

f x; y = f x; z , f y;z

ou encore
pour tous x, y , z 2 R.
C'est a dire

f x; z = f x;y + f y;z

f x; y = f x; z + f z;y


pour tous x, y , z 2 R, et donc f z;y  = ,f y;z . On pose f x; 0 = g x,
alors f 0;x = ,g x et
f x; y = f x; z + f z;y
= f x; 0 + f 0;y
= g x , g y  .
Reciproquement, f x;y  = g x , g y  ou g est une fonction arbitraire.

Alors f x; y  = f x;z  + f z;y , et f convient.
Dans le cas ou a + b = 1, b = 1 , a, et  s' crit

e
f x;y = af x; z + 1 , af y;z ,

et alors f x;x = f x;z  et donc pour tous x, y 2 R, f x;y  = f x;x.
Maintenant  donne
f x;x = af x; x + 1 , af y;y ,
et par cons quence
e

1 , af x;x = 1 , af y;y  .


Deux possibilit s se pr sentent. Soit a = 1 ou f x;x = f y;y  = f x; y ,
e
e
et f est constante. Si a = 1, b = 0, alors f x;y  = f x;z  pour tous x,

84

y, z 2 R. Donc f x;y = f x; x ind pendant de y. On a veri e que ces


e

fonctions conviennent.
En conclusion :
si a; b = 1; ,1, f x; y  = g x , g y  ou g : R ! R est arbitraire ;

si a + b 6= 1 et a; b 6= 1; ,1, f 0 ;
si a + b = 1 et a 6= 1 : f constante ;
si a; b = 1; 0 : f x; y  = g x pour tous x; y 2 R ou g : R ! R est

arbitraire.
9. On the plane there are given four distinct points A, B, C , D lying
in this order on a line g , at distances AB = a, BC = b, CD = c.
a Construct, whenever possible, a point P , not on g , such that the angles
APB, BPC , CPD are equal.
b Prove that a point P with the property as above exists if and only if the
following inequality holds: a + bb + c 4ac.
Solution by Michel Bataille, Rouen, France.
a If P is a solution, then the lines PB and PC are interior bisectors
PA = BA
in 4APC and 4BPD respectively. Hence we have:

PC
PB = CB and P is simultaneously on E = M : MA = a BC
and
1
PD CD
MC b and
MB b
E2 = M : MD = c .
In the general case where a = b, denoting by B 0 the harmonic conjugate
6
of B with respect to A and C , E1 is the circle with diameter BB 0 and, when
a = b, E1 is the perpendicular bisector of the segment AC . Similar results
hold for E2 .
Conversely, we may construct E1 and E2 and, assuming that they are
secant, choose for P one of their two distinct points of intersection symmetPA = BA , we deduce that PB is one of the bisectors
rical about g . From
PC BC
of APC , more precisely the interior bisector in 4APC since B is between
A and C . Hence APB = BPC . Similarly BPC = CPD and nally:
APB = BPC = CPD.
b The above construction provides a point P solution whenever E1 and E2
are secant. We rst examine the general case where a = b and b = c: E1 and
6
6
E2 are circles with centres I1, I2 and radii r1, r2 respectively. These circles
are secant if and only if:

jr1 , r2j

I1I2

r1 + r2
1
k
Let us denote by k the real number such that BI 1 = BC so that jkj = r1.
b

85

k + a BC and I C = b , k BC, and from the


1
b
b
ab , so that
2
Newton's relation, I1 B = I1 A  I1 C , we obtain easily k =
a,b
ab . Similarly: r = cb .
r1 = ja , bj
2
jc , bj
2 , ac
2
b
We also compute: I1 I2 =
BC so that I1I2 = jb bjbaj,bacj cj .
b , ab , c
, j ,
We may compute: I1 A = ,

The condition 1 may now be successively written:

jcja , bj , ajc , bj j jb2 , acj ajc , bj + cja , bj


a2c , b2 + c2a , b2 , 2acja , bj jc , bj b2 , ac2
a2c , b2 + c2 a , b2 + 2acja , bjjc , bj
jb2 , ac2 , a2c , b2 + c2a , b2j 2acja , bj jc , bj
ja , bj jc , bj jb2 + ba + c , acj 2acja , bj jc , bj
,2ac b2 + ba + c , ac 2ac
,ac b2 + ba + c 3ac .
Since b2 + ba + c is positive, the latter condition is equivalent to
b2 + ba + c 3ac or a + bb + c 4ac.
E1 and E2 are both lines when a = b = c, but in this case they are
strictly parallel so that no point P exists and the condition
a + bb + c 4ac is not true either.
Lastly, suppose for instance that E1 is a line and E2 is a circle that is,
a = b and b = c. Since E1 is perpendicular to g at B,2E1 and E2 are secant
6
b
if and only if I2 B r2 . We obtain easily: I2 B =
jc , bj and the condition
becomes: b c and the inequality a + bb + c 4ac reduces to b c
as well. The proof of b is now complete.

That completes our le of solutions for problems of the February 1998


number of the Corner. The Olympiad Season is nearly upon us. Send me your
national and regional Olympiads for use in the Corner. We also welcome your
nice solutions to problems that appear in the Corner.

86

BOOK REVIEWS
ALAN LAW

A Primer of Real Functions, by Ralph Boas Jr.,


published by the Mathematical Association of America, 1996,
ISBN 0-88385-029-X, softcover, 262+ pages, $32.95.
Reviewed by Murray Klamkin, University of Alberta.
This is the fourth edition of a popular classic Carus monograph which
has been revised, updated and augmented by the author's son Harold P.
Boas. The previous editions covered sets, metric spaces, continuous functions and di erentiable functions. This edition adds a chapter on measurable
sets and functions, the Lebesgue and Stieltjes integrals, and applications. The
new material is a rewrite of a draft left over by the author at his death. This
book can be likened to a sequence of lectures on a variety of topics selected
from the foundations of analysis and is done in a friendly and lively manner.
Mathematically Speaking: A Dictionary of Quotations, selected and arranged
by Carl C. Gaither and Alma E. Cavazos-Gaither,
published by Institute of Physics Publishing, 1998,
ISBN 0-7530-0503-7, softcover, 484+xiii pages, $39.00 US.
Reviewed by Bruce Shawyer, Memorial University of Newfoundland.
The book contains hundreds of quotations from hundreds of authors, as
well as many apocryphal quotations from persons unknown. The quotations
are grouped into 199 sets, ordered by topics, running from ABSTRACTION to
ZERO. Most sections are just a few pages long, except for the topics MATHEMATICIAN and MATHEMATICS, which have 29 and 80 pages respectively.
Also included is a complete bibliography of the source material plus
two excellent indices, the SUBJECT BY AUTHOR INDEX and the AUTHOR BY
SUBJECT INDEX.
The quotations vary from the profound to the witty. There are quotations from plays, and quite a few are in poetry, including several mnemonics
for  . Unfortunately, this reviewer's favourite is missing!
How I want a drink
3:1415
Alcoholic of course
926
After the heavy lectures
5358
Involving decimal fractions
979
Engineers can substitute quantum mechanics" for decimal fractions".
The authors quoted come from all di erent walks of life, from professional mathematicians and scientists to historians, journalists, philosophers,

87
poets, rap artists and writers. Some are famous, their names being almost
household words; others are much less known. As might be expected, by far
the majority of the quotations come from mathematicians and scientists.
To give a avour of this book, here is a selection of some of the shorter
quotations:
Some quotations from famous mathematicians
Richard Guy
Mathematics often owes more to those who ask questions than to
those who answer them.
Paul Halmos
The only way to learn mathematics is to do mathematics.
Leopold Kronecker
Number theorists are like lotus-eaters | having once tasted of this
food, they can never give it up.
George P
olya
Geometry is the art of correct reasoning on incorrect gures.
Some quotations from others
Ice-T
I write rhymes with addition and algebra, mental geometry.
G.K. Chesterton
Poets do not go mad; but chess-players do. Mathematicians go
mad, and cashiers; but creative artists very seldom.
John Updike
When you look into a mirror rorrim a otni kool uoy nehW
it is not yourself you see ees uoy esruoy ton si ti
but a kind of apish error rorre hsipa fo dnik a tub
posed in fearful symmetry yrtemmys lufraef ni desop
Mae West
A gure with curves always o ers a lot of interesting angles.
Unknown
Trigonometry is a sine of the times!
The book is a wonderful compendium and a great source of useful wisdom
for teachers of mathematics. It is very readable, and, once one has started
to read it, very di cult to put down.

88

THE SKOLIAD CORNER


No. 36

R.E. Woodrow
We begin with the problems of the Mini demi- nale 1996 of the Vingt
et uni me Olympiade Math matique Belge, organized by the Belgian Mathe
e
ematics Teachers Society. Twenty- ve of the problems are multiple choice.
The remaining ve require an integer answer between 0 and 999 inclusive.
Students are given 90 minutes. My thanks go to Ravi Vakil, Canadian Team
Leader to the International Mathematical Olympiad at Mumbai for collecting
the materials.


OLYMPIADE MATHEMATIQUE BELGE
Mini demi- nale 1996
Mercredi 6 mars 1996

1. Sans r eponse pr eformul ee | Quel est le nombre premier le plus


proche de 100 ?
2. Laquelle des propositions ci-dessous est la n egation de: Chaque
langue europ enne est parl e par l'un de nos guides au moins." ?
e
e
a Chacun de nos guides parle toutes les langues europ ennes.
e
b Chacun de nos guides parle une langue europ enne au moins.
e
c Aucun de nos guides ne parle aucune langue europ enne.
e
d L'un de nos guides ne parle aucune langue europ enne.
e
e L'une des langues europ ennes n'est parl e par aucun de nos guides.
e
e
3. Si P d esigne le p erim etre du triangle ABC et kABk, kAC k, kBC k
les longueures de ses c^ t s, laquelle des relations suivantes est correcte ?
oe
a kAB k = P=3
b kAB k  P=3
c kAB k  P=2
d kAB k + kAC k = 2P=3
e kAB k + kAC k  2P=3
4. Un parterre rectangulaire de 8 m sur 6 m est entour e ext erieurement
d'un sentier de 1; 5 m de large. Quelle est l'aire de ce sentier ?
a 23; 25 m2
b 37; 5 m2
c 42 m2
d 46; 5 m2
e 51 m2

89

5. Le graphique ci-dessous donne, pour les quatorze premiers jours de


sa vie, les gains de poids en grammes, et par rapport au poids a la naissance

d'un b b dont le poids a la naissance etait de 3; 250 kg.
e e


p g

400
200
0
,200

r
r r

r r

r r

2 4 r6 r 8 10 12 14 t jours

r r r

Quel etait son poids a une semaine ?




a 0; 100 kg b 3; 300 kg c 3; 350 kg d 3; 650 kg e 4; 250 kg
6. Si X , Y et Z sont les sommets d'un triangle, quel est le nombre de
parall logrammes admettant X , Y et Z pour sommets ?
e
a 1
b 2
c 3
d 4
e 6
7. Que vaut 3x2 , 7x2x3 , x2 + x , 2 ?
a 6x4 , 17x3 + 10x2 , 13x + 14.
b 6x5 , 17x4 + 10x3 , 13x2 + 14x.
c 6x5 + 17x4 + 10x3 + 13x2 + 14x.
d 6x6 , 17x5 + 10x4 , 13x3 + 14x2 .
e 6x6 , 3x4 , 14x3 + 10x2 , 7x + 8.
8. Sans r ponse pr eformul e | Une bille m etallique a la propri et e de
e
e
rebondir a une hauteur egale aux huit dixi mes de sa hauteur initiale; si elle


e
est l^ ch e d'une hauteur de 1; 25 m, quelle sera, en centim tres, la hauteur
a e
e
de son troisi me rebond ?
e
9. De combien augmente l'aire totale d'un cube lorsque la longueur de
chacune de ses ar^ tes augmente de 50 ?
e
a 50
b 125
c 225
d 237; 5
e 2500
10. Quatre sacs opaques contiennent :
Le sac A, une bille blanche et une bille rouge ;
Le sac B , deux billes blanches et deux billes rouges ;
Le sac C , deux billes blanches, une bille rouge et une bille noire ;
Le sac D, dix billes blanches et dix billes noires.
De quel sac faut-il tirer une bille au hasard pour avoir le plus de chances que
la bille tir e soit blanche ?
e
a Le sac A
b Le sac B
c Le sac C
d Le sac D
e Le choix est indi erent

90

11. Laquelle des


a

gures suivantes est le d veloppement d'un cube ?


e
b
c
d
e

12. Si la somme de trois nombres naturels est un nombre de deux


chi res, il est certain que :
a chacun des trois nombres est sup rieur a 10 ;
e

b deux des nombres, au moins, sont inf rieurs a 50 ;
e

c aucun des trois nombres n'est sup rieur a 50 ;
e

d les trois nombres sont di erents ;

e le produit des trois nombres est inf rieur a 35 000.
e

13. Avant son d epart en vacances, une personne a achet e 3000 francs
fran ais pour 18 270 francs belges. Cette personne, en France, a d^ changer
c
u
a nouveau de l'argent: pour 10 000 francs belges, elle a re u 1600 francs

c
fran ais. Si elle avait achet en Belgique, avant son d part, tout l'argent
c
e
e
fran ais dont elle a eu besoin,
c
a elle aurait gagn 480 francs belges ;
e
b elle aurait gagn 256 francs belges ;
e
c cela serait revenu au m^ me ;
e
d elle aurait perdu 256 francs belges ;
e elle aurait perdu 480 francs belges.

14. Une personne a achet e des timbres a 3 F et des timbres a 5 F



pour un total de 100 F exactement. Parmi les suivants, quel est le nombre
de timbres a 5 F qu'elle ne peut pas avoir achet ?

e
a 5
b 8
c 9
d 11
e 17

15. Un cycliste monte une c^ te a la vitesse moyenne de 12 km h, pour


o
la redescendre ensuite a la vitesse moyenne de 48 km h. Si la mont e a dur

e
e
22 min 30 s de plus que la descente, quelle est la longueur de cette c^ te ?
o
a 6 km
b 8 km
c 12 km
d 13; 5 km
e 15 km

91

16.

A
D
F

Voici un hexagone ABCDEF dans


lequel les c^ t s oppos s sont paroe
e
all les et de m^ me longueur. Le trie
e
angle ABC est n cessairement ape
pliqu sur le triangle DEF par
e

a une translation


b une sym trie orthogonale
e
c une sym trie central
e
d une rotation de 90
e aucune des transformations pr c dentes
ee
17. Quel est le nombre maximum de points communs a un cercle et
au bord d'un losange ?
a 2
b 4
c 6
d 8
e 10
18. Laquelle des a rmations suivantes est vraie ?
a Il existe des carr s qui ne sont pas des rectangles.
e
b Un carr n'est jamais un rectangle.
e
c Tout parall logramme est un losange.
e
d Tout losange est un parall logramme.
e
e Certains rectangles ne sont pas des parall logrammes.
e
19. Dix nombres sont tous inf erieurs a 20 et leur moyenne arithm etique
vaut 18. Laquelle des a rmations suivantes est certainement correcte a pro
pos de ces nombres ?
a L'un d'entre eux, au moins, est egal a 18.

b Un nombre pair d'entre eux sont egaux a 18.


c Un nombre impair d'entre eux sont egaux a 18.


d Ils sont tous sup rieus a 16.
e

e Ils sont tous positifs.
20. Sans r ponse pr formul ee | Les roues avant d'un wagonnet ont
e
e
7 cm de circonf rence et les roues arri re ont 9 cm de circonf rence. Lorsque
e
e
e
les roues avant ont fait 10 tours de plus que les roues arri re, de combien de
e
centim tres a avanc le wagonnet ?
e
e
21. La somme de deux nombres premiers est toujours
a un nombre pair ;
b un nombre impair ;
c un nombre premier ;
d strictement sup rieure a 3 ;
e

e inf rieure a 1000.
e

22. Le poids moyen des 30 e l eves d'une classe est de 47 kg; si chacun
de ces el ves grossit de 3 kg, de combien augmentera le poids moyen ?
e
a 0; 1 kg
b 2 kg
c 3 kg
d 90 kg
e Une autre valeur

92

23.

10
20

10

Dans le train d'engrenages


repr sent ci-contre, lorsque la
e e
roue R fait un tour dans le sens des
aiguilles d'une montre, de combien
tourne la roue S ? Les nombres
indiqu s donnent le nombre de
e
dents de chaque engrenage. Deux
cercles concentriques repr sentent
e
deux roues solidaires du m^ me axe.
e

a D'un tour dans le sens des aiguilles d'une montre.


b De 2 tours dans le sens des aiguilles d'une montre.
c De 4 tours dans le sens des aiguilles d'une montre.
d De 2 tours dans le sens oppos a celui des aiguilles d'une montre.
e
e De 4 tours dans le sens oppos a celui des aiguilles d'une montre.
e

24. Sans r ponse pr eformul e | Pour un certain nombre naturel n,


e
e

2n + 3 est un diviseur de 6n + 43; que vaut n ?

25. Un marchand de jouets a achet e un lot de 100 oursons en peluche


qui valaient au total 21 000 F, mais le grossiste lui a consenti une remise de
10. Ce marchand souhaite qu'en accordant une remise de 25 sur le prix
de vente a ch , il lui reste encore un b n ce egal a 30 du prix de vente
e
e e
r el. A combien doit-il a cher l'ourson ?
e
a 341; 25 F

alors,

b 360 F

c 400 F

d 34 125 F

e Une autre r ponse


e

26. Soit x et y deux demi-droites contenues dans une m^ me droits ;


e

a il existe n cessairement une sym trie centrale appliquant x sur y ;


e
e
b il existe n cessairement une sym trie orthogonale appliquant x sur y ;
e
e
c il existe n cessairement une translation appliquant x sur y ;
e
d il existe n cessairement une rotation appliquant x sur y ;
e
e aucune des propositions pr c dentes n'est vraie.
ee

93

27. Un avion vole a la vitesse de 400 km h par rapport a l'air. Pendant




son voyage aller, il a de face un vent de 40 km h; au retour, il a le m^ me vent
e
dans le dos. Quelle est sa vitesse moyenne sur l'ensemble du trajet ?
a 390 km h b 396 km h c 398 km h d 400 km h e 410 km h
Last issue we gave the problems of the Old Mutual Mathematical
Olympiad 1992. Thanks go to John Grant McLoughlin, Faculty of Education, Memorial University, St. John's, Newfoundland for the problem set
and solutions. Here are the answers:
1.
5.
9.
13.
17.

c
a
a
e
a

2.
6.
10.
14.
18.

b
c
b
a
e

3.
7.
11.
15.
19.

d
c
e
c
a

4.
8.
12.
16.
20.

c
d
b
e
e

That completes the Skoliad Corner for this issue. Please send me contest materials and suggestions for other features of the Corner.

XXXX
Y

YY

Y
X

XX

Y
X

YYY
YY
XXX
XX
YYY YY
Y

XX

YY

XXXXX

YYYY

94

MATHEMATICAL MAYHEM
Mathematical Mayhem began in 1988 as a Mathematical Journal for and by
High School and University Students. It continues, with the same emphasis,
as an integral part of Crux Mathematicorum with Mathematical Mayhem.
All material intended for inclusion in this section should be sent to the
Mayhem Editor, Naoki Sato, Department of Mathematics, Yale University,
PO Box 208283 Yale Station, New Haven, CT 06520 8283 USA. The electronic
address is still
mayhem@math.toronto.edu
The Assistant Mayhem Editor is Cyrus Hsia University of Toronto.
The rest of the sta consists of Adrian Chan Upper Canada College, Jimmy
Chui Earl Haig Secondary School, David Savitt Harvard University and
Wai Ling Yee University of Waterloo.

Shreds and Slices

Non-mathematical Problem

In Problem Book for First Year Calculus, by George W. Bluman, SpringerVerlag, Problem 13.7 asks: Choose a non-mathematician:
a John von Neumann.
b Mick Jagger.
c Georg Cantor.
d Pablo Casals.
e Stanley I. Grossman.
f Ren Descartes.
e
g Guy La eur.
The answer at the back of the book? Possibly b, d, g."

Awaiting a Combinatorial Proof

Find a combinatorial proof of the following identity:




r
n , r n + r , 1 n = n n +2rr , 1 2rr :
r

The individual with the rst correct solution that is strictly combinatorial will
get a free book prize and the solution published here.
Reference
Tucker, Alan. Applied Combinatorics. John Wiley & Sons, Inc. Toronto.
1995 pp. 221.

95

Discovering the Human Calculator in You


Richard Hoshino

student, University of Waterloo


In his Oscar-winning role in the movie Rain Man, Dustin Ho man plays
an idiot savant who can perform complex calculations instantly in his head.
Like the Rain Man", various people have displayed their outstanding capacity for mental arithmetic on TV, and many others have written books teaching these powerful techniques. However, hardly any have ever ventured to
justify the validity of these algorithms, as the mathematics involved is surprisingly elementary. In this article, we detail some of the famous tricks that
the human calculators" have used over the years, explain why these methods work, and you will see that, with a little practice, you too can be a human
calculator.
Trick 1. Squaring two-digit numbers ending in 5.
To square any two-digit number that ends in 5, add one to the rst digit
and multiply that sum by the rst digit. This will be the rst two digits of
the answer. The last two digits will always be 25.
For example, 852 = 7225 since 8  8 + 1 = 8  9 = 72, and likewise,
2 = 625 since 2  3 = 6. We can extend this to larger numbers, for
25
example, 1952 = 38025, since 19  20 = 380.
If you are wondering why this method works, a little algebra will quickly
convince you:

10A + 52 = 100A2 + 100A + 25 = 100AA + 1 + 25 .


Thus, the rst two digits will be AA + 1, and the last two digits will
be 25.
Let us take this idea one step further. Let us multiply pairs of two-digit
numbers whose tens digits are the same, and whose units digits sum to ten.
For example, 37  33 = 1221, 36  34 = 1224, and 98  92 = 9016.
Do you see the pattern? Like in the case above, the rst two digits of
the answer are determined in the same way. But what about the last two
digits? Do you see how they are obtained? If so, use a little algebra and
convince yourself that it always works.
Copyright c 1999 Canadian Mathematical Society

96
Trick 2. Squaring any two-digit number.
Take any two-digit number n. Now we know that

n2 = n2 , d2  + d2 = n , dn + d + d2 ,
so let us try to nd a value of d so that the product n , dn + d can be
easily calculated. Consider the multiple of 10 that is closest to n, and let the
di erence between the number and this multiple of 10 be d. For example, if
we take n = 87, then the multiple of 10 that is closest to 87 is 90, and since
90 , 87 = 3, we have d = 3. Similarly, if n = 94, we have d = 4.
If we perform this calculation for any integer n, then one of n , d or
n+d will be a multiple of 10, and the calculation becomes signi cantly easier.
The following examples illustrate this technique:

872 = 87 + 387 , 3 + 32 = 90  84 + 9 = 7569 ,


292 = 29 + 129 , 1 + 12 = 30  28 + 1 = 841 ,
962 = 96 + 496 , 4 + 42 = 100  92 + 16 = 9216 .
Use this technique to compute the following: 372 , 522, 19992.

Trick 3. The Calendar Trick.


One of the more interesting demonstrations performed by mathemagicians" is the calendar trick. Namely, an audience member calls out her
birthday, or some historical date, and the human calculator is able to tell her
what day of the week that event took place.
The rst thing to do is to memorize the following table:
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

It appears challenging to remember this, but there is an interesting pattern here. Reading the row of numbers from left to right in threes, we have
144, 025, 036 and 146. Notice that 144, 025 and 036 are perfect squares,
and the last number is just 2 more than the rst perfect square, 144. This
should make the memorization easier.
Let Y be the last two digits of the year in question. Let D be the day
we are searching for, and let M be the integer that corresponds to the month
in the above table. Thus, if we are searching for July 25th, 1978, Y would
be 78, D would be 25, and M would be 0, since July corresponds to 0.
Compute the value of

Y +D+M,
Y+ 4


and divide that sum by 7. Whatever remainder you get corresponds to the day
of the week you are seeking, namely 0 is Saturday, 1 is Sunday,
2 is Monday, 3 is Tuesday, 4 is Wednesday, 5 is Thursday, and 6 is Friday.

97
A small note to remember. If the year is a leap year, and the month is
January or February, you must subtract 1 from the total. This is due to the
fact that the extra day in a leap year occurs on February 29th, and so if the
day you are searching is before that, then the formula is o by one day.
Let us look at a historical date in the 20th century. The famous stock
market crash of 1929 occurred on October 29th , so let us use our formula to
determine what day of the week Black Tuesday" occurred.
We have Y = 29, M = 1, and D = 29. Hence b 29 c = 7, and our
4
sum is 29 + 7 + 29 + 1 = 66. Dividing this number by 7, we nd that the
remainder is 3. We conclude that October 29th, 1929 was indeed a Tuesday.
Unfortunately, this formula only works with dates in the 1900's, because in the Gregorian calendar, not all years that are divisible by 4 are leap
years. For example, 1800 and 1900 are not leap years, but 2000 is. And
thus, we must alter our formula to compensate for this. To calculate dates
in the 1800's, use the same formula, but go forward two days in the week.
To calculate dates in the 2000's, go one day back. If we use our formula, we
nd that January 1st, 2000 is a Sunday remember, 2000 is a leap year!. Go
back one day, because it really is a Saturday.
In the past, it was believed that a year had precisely 365:25 days, and
so we compensated for the extra quarter day by adding February 29th to our
calendar once every four years. Unfortunately, a year has 365:2422 days, so
we cannot add an extra day exactly once every four years. It would be nice
however if we could, for then this formula would always hold.
Using this method, determine what day of the week you were born on.
Trick 4. Extracting Cube Roots.
We now detail the method for determining the cube roots of all perfect
cubes under one billion.
First, you must rst learn the cubes of the integers 0 through 9.

n 0 1 2 3 4 5 6 7 8 9
n3 0 1 8 27 64 125 216 343 512 729

Let us rst nd the cube root of numbers that are below one million.
Hence, the cube root will be at most 99. Say we want to nd the cube root of
314; 432. We separate the number into two parts, separated by the comma.
Thus, 314 is the rst part, and 432 is the second part. The desired cube root
has two digits. We will use the rst part to get the rst digit, and we will use
the second part to get the second digit.
Take the rst part and determine where it lies in the table of cubes. In
our example, 314 lies between 216 and 343. Thus, 216; 000 314; 432
343; 000, which implies that the desired cube root lies between 60 and 70,
since 603 = 216; 000 and 703 = 343; 000. Hence, it follows that the rst
digit of our cube root must be 6.

98
Now we determine the second digit. If we look at the table of cubes,
notice that each cube ends in a di erent digit. So if a certain cube ends in 2,
we know that its cube root must end in 8, because 8 is the only digit whose
cube ends in 2. Since 432 ends in 2, the second digit of the cube root must
be 8. Thus, the desired cube root is 68.
Before, you go further, determine the cube roots of the following numbers in your head: 157; 464; 185; 193; 778; 688; 12; 167.
With a little practice, you will nd that it is faster to do this exercise in
your head rather than punching it in your calculator!
So that we do not confuse the digits in the following example, let us
introduce some variables. Let n = 100p +10q + r3 ; that is n is the perfect
cube, and 100p +10q + r is the cube root of n, where p, q and r are its digits.
To do this trick, once again, you need to memorize a small table:

A 1 2 3 4 5 6 7 8 9 10
B 1 7 9 5 3 8 6 2 4 10
Just keep repeating one-seven-nine- ve-three, eight-six-two-four-ten", to
remember the right order in the B row. It might help to notice that the rst
ve entries are odd, and the last ve entries are even.
Now, let us move on to the cubes of three-digit numbers. Let us say
we wanted to nd the cube root of n = 101; 847; 563. First, let us separate
this large number into three smaller ones, separated at the commas.
As we did before, we look at the rst part, 101, to determine what the
rst digit of the cube root is. Since 101 lies between 64 = 43 and 125 = 53 ,
we conclude that the p = 4. The last part is 563, which ends in a 3, and since
73 = 343, we see that r = 7. Hence, we determine the rst and last digits
the same way as we did before with the smaller numbers.
Now take n, and add and subtract the digits of this number in alternating fashion starting from the right. Thus, we compute
3 , 6 + 5 , 7 + 4 , 8 + 1 , 0 + 1 = ,7.
Now repeatedly add or subtract 11 to this number until we get a number
between 0 and 10 inclusive more formally, we say we take this number
modulo 11. Hence, ,7 becomes 4. Let this number be A. If you have
ever seen the test for divisibility by 11, you will see that n A mod 11.
Now take the number A, and look in the above table to determine
the corresponding number B . We see that A = 4 corresponds to
B = 5; thus we have B = 5. Finally, our second digit q is the value of
p + r , B mod 11; that is we add or subtract 11 until we get an integer
between 0 and 10 inclusive, and this will be our digit q . Thus, in our case,
p + r , B = 4 + 7 , 5 = 6, and so we have shown that q = 6.
Therefore, p = 4, q = 6 and r = 7; thus the cube root of
n = 101; 847; 563 is 467. Checking, we see that this is correct.

99
Let us summarize the algorithm.
i Use the rst part of n to determine the rst digit of the cube root. Call
this digit p.
ii Use the last digit of n to determine the last digit of the cube root. Call
this digit r.
iii Take the alternating sum of n to determine A, where A is between 0
and 10. That is, nd the A for which n A mod 11.
iv Use the table to nd the number B that corresponds to A.
v Determine the sum p + r , B , and reduce it modulo 11. This is q .
vi The number with rst digit p, second digit q and third digit r is your
desired cube root.
Now try to nd the cube roots of the following numbers. Note, you can
use a pencil and paper, but you are not allowed to use a calculator!
a 17; 173; 512, b 1; 860; 867, c 758; 550; 528,
d 84; 604; 519, e 170; 953; 875.
Let us now justify why this algorithm works. First of all, take each
value B in the table, and compute B 3 mod 11. You will nd that we will
get the corresponding value of A in each case. That is how the numbers are
determined. Note that there is a bijection between the elements of the two
rows in the table. If this were not the case, then this algorithm would not
work. We have

n = 100p + 10q + r3


p , q + r3
A mod 11 .
And from the table, this implies that p , q + r B mod 11; that
is q p + r , B  mod 11.
Therefore, q is uniquely determined, as are p and r from before.

With a little practice, you will become comfortable with all of the methods described in this article. Maybe you will even be able to come up with
better ways to perform the tricks described. Nevertheless, through practice
and perseverance, you too can be a human calculator.

100

Mayhem Problems
The Mayhem Problems editors are:
Adrian Chan Mayhem High School Problems Editor,
Cyrus Hsia Mayhem Advanced Problems Editor,
David Savitt Mayhem Challenge Board Problems Editor.
Note that all correspondence should be sent to the appropriate editor |
see the relevant section. In this issue, you will nd only solutions | the
next issue will feature only problems.
We warmly welcome proposals for problems and solutions. With the
new schedule of eight issues per year, we request that solutions from the last
issue be submitted in time for issue 2 of 2000.

High School Solutions


Editor: Adrian Chan, 229 Old Yonge Street, Toronto, Ontario, Canada.
M2P 1R5 all238@ipoline.com

H223. In each of the following alphametics, each letter in the addition represents a unique digit:
+

1 9 9 7

1 9 9 8

O L D and +
O L D .
Y E A R
Y E A R

For each alphametic, nd a solution, or prove that a solution does not exist.
Solution. First, we show that the second alphametic has no solution.
If either L or D is greater than 1, then there will be a carry involved in each
step of the addition. Most importantly, from the third column, we obtain
that 1 + 9 + O E mod 10, which implies that O E mod 10.
Thus, O and E represent the same digit, which we cannot have. Hence,
a solution must satisfy LD = 01; that is L = 0 and D = 1. But this leads to

1 9 9 8
O 0 1,
Y E 9 9

so A = R = 9, which we cannot have either. Therefore, there is no solution


to this alphametic.
A solution to the rst alphametic must satisfy L = 0 and D = 1, or
L = 0 and D = 2 by the same reasoning as above. Setting L = 0 and

101

D = 1 leads to A = R = 9, but setting L = 0 and D = 1 leads to several


possibilities, for example

1 9 9 7
4 0 1 .
2 3 9 8

H224. Let ABCD be a square. Construct equilateral triangles APB,


BQC , CRD, and DSA, where P , Q, R, and S are points outside of the
square.
a Prove that PQRS is a square.
b Determine the ratio PQ=AB . See how many ways you can solve this!
Solution.
P

60

60

S
D

R
a In the gure, PAS = PBQ = QCR = RDS = 150 , and
AP = AS = BP = BQ = CQ = CR = DR = DS, so triangles
PAS, PBQ, QCR, and RDS are congruent. Hence, PQ = QR = RS =
SP . Also, PAS = 150 , so SPA = PSA = 15 , and SPQ =
APB + BPQ + BPQ = 15 + 60 + 15 = 90 , and by symmetry,
PQR = QRS = RSP = 90 as well, so PQRS is a square.
b By the Cosine Law,

PQ2 = PB2 + BQ2 , 2PB  PQ cos 150


= 2AB 2 , 2AB 2  cos 150 = 2AB 2 1 + cos 30 

!
PQ2 = 2 1 + p3 = 2 + p3 = 4 + 2p3
= AB2
2
2
s
PQ = 4 + 2p3 = 1 + p3 = p2 + p6 .
= AB
p2
2
2

102

H225. Consider a row of ve chairs, numbered 1, 2, 3, 4, and 5. You


are originally sitting on 1. On each move, you must stand up and sit down
on an adjacent chair. Make 19 moves, then take away chairs 1 and 5. Then
make another 97 moves, with the three remaining chairs. No matter how
the moves are made, you will always end up on chair 3. Why is this the case?
Solution. Say that chairs 1, 3, and 5 are odd chairs and that chairs 2
and 4 are even chairs. After 1 move, we are on an even chair. After 2 moves,
we are on an odd chair. By a simple parity argument, after 19 moves we
must be on an even chair. Thus, we take away chairs 1 and 5, and are left
with two even chairs and one odd chair.
By parity again, after 97 moves, we will be on an odd chair, namely
chair 3. Hence, we will always end up on chair 3.
H226. The smallest multiple of 1998 that consists of only the digits
0 and 9 is 9990.
a What is the smallest multiple of 1998 that consists of only the digits 0
and 3?
b What is the smallest multiple of 1998 that consists of only the digits 0
and 1?
Solution. a Let N be the smallest multiple of 1998 that only consists
of the digits 0 and 3. Since 1998 = 6  333, we have that 333 divides N=6.
By de nition, N is of the form

N = 3  10a + 3  10a +    + 3  10an


for some integers a1 a2    an 0. Then,
N = 5  10a ,1 + 5  10a ,1 +    + 5  10an,1 ,
1

which implies that N=6 is an integer that consists of only the digits 0 and 5.
Since 9 divides 333, 9 must divide N=6. By a well known divisibility test, the
number of 5's in N=6 must be a multiple of 9. Hence, N=6  555 555 555,
or N  3 333 333 330.
Checking, we nd that 1998 does indeed divide 3 333 333 330, so this
is the multiple of 1998 we seek.
b Let N be the smallest multiple of 1998 that consists of only the digits 0
and 1. Then N is of the form

N = 10a + 10a +    + 10an


for some integers a1 a2    an 0. The last digit of N must
be a 0. Note that 1998 divides N if and only if 999 divides N=2, which
1

103
in turn occurs if and only if 999 divides N=10, so henceforth we consider
N=10 = 10a ,1 + 10a ,1 +    + 10an,1. Let bi = ai , 1 for all i.
Let a, b, and c be the number of exponents bi which are congruent to
0, 1, and 2 modulo 3 respectively. Then
1

N = 10a ,1 + 10a ,1 +    + 10an,1


10
= 10b + 10b +    + 10bn
a + 10b + 100c
0 mod 999 ,
1

since 103 1 mod 999. Note that a + 10b + 100c 0 mod 999 implies
that 10a + 100b + c 0 and 100a + b + 10c 0, which we obtain by
multiplying by 10 and 100 respectively. We consider possible values of a, b,
and c.
Suppose that c  10. Since a + 10b + 100c a + 10b + 100c , 10 +
1000 a + 1 + 10b + 100c , 10 mod 999, the triple a + 1; b; c , 10
is also a solution. Similarly, if b  10 or a  10, then we can obtain another
solution by using the two congruences derived above. Hence, by using this
reduction, for any given solution, we can obtain a solution where a, b, and c
are all at most 9. There is no danger of being caught in a cycle, since each
application of the reduction decreases the sum a + b + c by 9.
Now consider such a reduced solution. Each of a, b, and c is at most 9,
but if any is less than 9, then it is apparent that a + 10b + 100c 999, so
the congruence a + 10b + 100c 0 mod 999 cannot hold. Therefore, the
only solution with all of a, b, and c at most 9 is a = b = c = 9. This leads
to the number

N = 111:::1 .
10 | z
27 1's

We have ruled out the case a + b + c 27, and all values such that
a + b + c  27 must lead to a number with at least 27 digits. Therefore, the
desired minimal multiple is

N = 111z 1 0 .
| :::
27 1's

104

Advanced Solutions
Editor: Donny Cheung, c o Conrad Grebel College, University of Waterloo, Waterloo, Ontario, Canada. N2L 3G6 dccheung@uwaterloo.ca

A209. Is there an in nite number of squares among the triangular


numbers?
Solution.
The answer is yes. The problem reduces to a special case of Pell's equation. Pell's equation is the name given to diophantine equations of the form
x2 , dy2 = 1, where d is a positive, non-square integer.
We want to show that there is an in nite number of integers n such
that the nth triangular number nn , 1=2 is a perfect square. Assume that
nn , 1=2 = a2, or2equivalently nn , 1 = 22a2. By completing the square,
we obtain 2n , 1 , 8a2 = 1, or 2n , 1 , 22a2 = 1.
In other words, the problem reduces to showing that there is an in nite
number of pairs of integer solutions to the equation x2 , 2y 2 = 1. Here
x = 2n , 1 and y = 2a. This is a particular case of Pell's equation, which
has in nitely many solutions. First, note that x0 ; y0 = 3; 2 is a speci c
solution. Then note that for any solution x; y , 3x + 4y; 2x + 3y  is also
a solution:
3x + 4y 2 , 22x + 3y 2 = 9x2 + 24xy + 16y2 , 8x2 , 24xy , 18y 2
= x2 , 2y 2 = 1 .
Furthermore, if x is odd and y is even, then 3x +4y is odd and 2x +3y
is even. In this way, we can generate an in nite number of solutions to
x2 , 2y2 = 1 starting from 3; 2. Thus, setting n = xth 1=2, we nd that
+
there is an in nite number of integers n such that the n triangular number
is a perfect square.
A210. Proposed by Naoki Sato.
Let P be a point inside circle C . Find the locus of the centres of all
circles ! which pass through P and are tangent to C .
Solution.
Let the centre of circle C be O and its radius R. We claim that the locus
is an ellipse with foci at P and O, with semi-major axis length R=2. To see
this, let D be a circle passing through P and tangent to the circle C as shown.
Let Q be the centre of this circle.

105

D
r

C
r

qP
q

Q O

To show that Q lies on the ellipse with foci P and O, it is su cient to


show that PQ + QO is a constant. Let the radius of circle D be r. Since the
two circles C and D are tangent, the common tangent point is collinear with
the centres O and Q as shown. Then OQ = R , r, and PQ is the radius of
the circle. Therefore, PQ + QO = r + R , r = R. Thus, the locus is an
ellipse with foci at P and O. Further, the constant is R, so the length of the
ellipse's semi-major axis is R=2.
A211. Do there exist a convex polyhedron and a plane, not passing
through any of its vertices, and intersecting more than 2=3 of all of the edges
of the polyhedron?
Polish Mathematical Olympiad, rst round
Solution.
We rst prove a lemma.
Lemma. For any polyhedron with E edges and F faces, we have the
inequality 2E  3F .
Proof. Each face has at least three edges, so there is a total of at least
3F edges, counting each edge twice. Each edge is counted once for the two
faces it attaches. The total number of edges is given by E , so 2E  3F .
Suppose on the contrary that there are a convex polyhedron and a plane,
not passing through any of its vertices, and intersecting more than 2=3 of all
the edges. Let E , F , and V denote the number of edges, faces, and vertices of
the polyhedron respectively. The plane cuts the polyhedron into a polygonal
region. Suppose the polygon formed in this way has n vertices, and hence
n sides. Each vertex is the intersection of an edge of the polyhedron with
the plane. By assumption, n 2E=3. Now each edge of the polygon is the
intersection of a face of the polyhedron with the plane. Thus, the number of
faces of the polyhedron is at least n. That is, F  n.
Together with the lemma, we obtain that 2E=3  F  n 2E=3,
which is a contradiction.

106

Challenge Board Solutions


Editor: David Savitt, Department of Mathematics, Harvard University,
1 Oxford Street, Cambridge, MA, USA 02138 dsavitt@math.harvard.edu
This issue, we will be delaying the challenge board solutions until a
later issue. The observant reader will have noticed that the schedule for the
high school and advanced sections have been asynchronous with that of the
challenge board section; this will x this problem.
Note: Although the editor listed, as the proposer for C83 CRUX with
MAYHEM, Vol. 25, Issue 1, the person from whom he heard the problem,
the editor is aware that C83 is an old question. In fact, a version of the
problem has appeared in a competition as recently as 1996, when it was used
in a Romanian IMO team selection contest. Thanks to Mohammed Aassila
for bringing this to our attention.

Problem of the Month


Jimmy Chui, student, Earl Haig S.S.
that

Problem. Let a, b, and c be the lengths of the sides of a triangle. Prove


p
p
p
a + b , c + b + c , a + c + a , b  pa + b + pc ,

and determine when equality occurs.


1996 APMO, Problem 5
Solution. Let s = a + b + c=2, the semi-perimeter, and let x = s , a,
y = s , b, and z = s , c. Then a = y + z, b = z + x, and c = x + y. Note
that x, y , and z are all positive, since x = b + c , a=2, etc., and a, b, and c
are the lengths of the sides of a triangle. This is known as the infamous Ravi
Substitution Ed. at least in Canadian IMO circles . Hence, the inequality
is equivalent to

p2x + p2y + p2z  pz + x + px + y + py + z .

Recall that the Arithmetic Mean Quadratic Mean AM QM inequality
states that
s

a+b 
2

a2 + b2
2

107
for all a, b  0. Then, we have that

p2x + p2y + p2z = p2x + p2y + p2y + p2z + p2z + p2x


2
2 r
2
r
r
2x + 2y + 2y + 2z + 2z + 2x AM QM

2 p
2p
2
= pz + x + x + y + y + z :
Equality holds if and only if x = y = z ; that is, if and only if a = b = c.

When dealing with an expression involving the lengths of the sides of a


triangle, implementing the Ravi Substitution will often alter the expression
to a more manageable form.

J.I.R. McKnight Problems Contest 1987


1. a Three people, Aretha, Bob, and Chai, throw dice upon the condition that the one who has the lowest result shall give each of
the others the sum of money each of the two winners has already.
Aretha loses rst, Bob loses second, and Chai loses the third game.
They discovered that each nished with the same amount of money.
Express the amount of money that each one had at the beginning
in terms of the amount that each had at the end of the third game.
b Find all integer solutions for x, y , and z :
xy + z = 32 ,
yx + z = 65 ,
zx + y = 77 .
2. a Find the sum of the rst 22 terms of the geometric series having
p
rst term i and ratio 1 + i, where i = ,1. Give your answer in
the form a + bi, where a, b 2 R.
b A three-dimensional gure is de ned by the equation
4jxj + 3jy j + 6jz j = 12 .
Identify this gure and determine its volume.
3. A sequence of integers is de ned by the following recursion: x1 = 2,
x2 = 5, and xk = xk,1 + 2xk,2 for k 2. Prove that
n,1 +
n
xn = 7  2 3 ,1 .

108
4. The screen of a drive-in theatre is p units tall and is situated on a hill
q units high. A car is situated a distance x units from the screen such
that the angle subtended by the screen is a maximum. Show that the
p
maximum value of occurs when x = q p + q .

6
p

6 ?
?

5. The lengths of the sides of a triangle are 8, 8, and 11. Find the length
of one of the angle trisectors drawn to the longest side.
6. Consider the set of odd numbers f1, 3, 5, : : : , 101g.
a How many combinations of two distinct numbers can be formed
from this set?
b Determine the sum of the products of the pairs in a.
7. For any convex quadrilateral ABCD, the diagonals AC and BD intersect at E . The centroids of triangles ABE , BCE , CDE , and DAE are
P , Q, R, and S respectively.
a Prove that PQRS is a parallelogram.
b Find the ratio of the area of the parallelogram PQRS to the original quadrilateral.
Do question 8 or 9:
8. Seventeen dots are arranged so that no three are collinear. Each pair of
dots is connected by a line segment which may be drawn using one of
three colours. Prove that there are at least three points connected to
each other with the same colour.
9. Prove the following theorem: If the bisectors of a pair of opposite exterior angles of a cyclic quadrilateral are parallel, then the angles at the
other two vertices are right angles.

109

Swedish Mathematics Olympiad 1985


Final Round
4. The polynomial px of degree n has real coe cients, and px  0 for
all x. Show that
px + p0x + p00x +    + pnx  0 .
Iran.
have

Solution by Hadi Salmasian, Sharif University of Technology, Tehran,


Put x = px + p0 x +    + pnx. Since pn+1x = 0, we

x = px + 0 x .


Now, de ne  x = xe,x . Then
exx = px + ex x + ex0x ,

1

which is a result of 1. This gives us the following equation:


0x = ,pxe,x ,
and obviously, for all x 2 R,  0 x  0, so  x is a decreasing function.
On the other hand, the degrees of p0 , p00 , : : : , are less than n, and the leading coe cient in both polynomials px and x are equal. The following
lemma is easy to prove:
Lemma. For a non-zero polynomial px, the following two are equivalent:
i limx!+1 px = +1.
ii The leading coe cient of px is positive.
Using the lemma, we nd that x  0 for su ciently large values
of x, and the same is true for  x. But  x is a decreasing function, so
x  0 for all x. The same is also true for x.

110

PROBLEMS
Problem proposals and solutions should be sent to Bruce Shawyer, Department of Mathematics and Statistics, Memorial University of Newfoundland, St. John's, Newfoundland, Canada. A1C 5S7. Proposals should be accompanied by a solution, together with references and other insights which
are likely to be of help to the editor. When a submission is submitted without a solution, the proposer must include su cient information on why a
solution is likely. An asterisk ? after a number indicates that a problem
was submitted without a solution.
In particular, original problems are solicited. However, other interesting problems may also be acceptable provided that they are not too well
known, and references are given as to their provenance. Ordinarily, if the
originator of a problem can be located, it should not be submitted without
the originator's permission.
To facilitate their consideration, please send your proposals and solutions on signed and separate standard 8 1 "11" or A4 sheets of paper.
2
These may be typewritten or neatly hand-written, and should be mailed to
the Editor-in-Chief, to arrive no later than 1 October 1999. They may also
be sent by email to crux-editors@cms.math.ca. It would be appreciated if
A
email proposals and solutions were written in LTEX. Graphics les should
be in epic format, or encapsulated postscript. Solutions received after the
above date will also be considered if there is su cient time before the date
of publication. Please note that we do not accept submissions sent by FAX.

2414. Proposed by Wu Wei Chao, Guang Zhou Normal University,


Guang Zhou City, Guang Dong Province, China, and Edward T.H. Wang, Wilfrid Laurier University, Waterloo, Ontario.
For 1 x  e  y or e  x y , prove that xx y xy
xyx yx.
2415. Proposed by Paul Yiu, Florida Atlantic University, Boca Raton, Florida, USA.
Given a point Z on a line segment AB , nd a Euclidean construction of
a right-angled triangle ABC whose incircle touches hypotenuse AB at Z .
2416. Proposed by V aclav Konecn y, Ferris State University, Big
Rapids, Michigan, USA.
Given 4ABC , where C is an obtuse angle, suppose that M is the
mid-point of BC and that the circle with centre A and radius AM meets
BC again at D. Assume also that MD = AB. The circle, ,, with centre
M and radius MB meets AB at E. Let H be the foot of the perpendicular
from A to BC extended. Suppose that AC and EH intersect at I .
Find the angles IAH and AHI as function of ABC .
This proposal was inspired by problem 2316.

111
UK.

2417.

Proposed by Christopher J. Bradley, Clifton College, Bristol,

2418.

Proposed by Christopher J. Bradley, Clifton College, Bristol,

In 4ABC , with AB = AC , the internal and external bisectors of


6
BAC meet the circumcircle of 4ABC again in L and M respectively. The
points L0 and M 0 lie on the extensions of AL and AM respectively, and
satisfy AL = LL0 and AM = MM 0 . The circles ALM 0 and AL0 M meet
again at P .
Prove that AP k BC .
UK.

In 4ABC , the lengths of the sides BC , CA, AB are 1998, 2000, 2002
respectively.
Prove that there exists exactly one point P distinct from A and B  on
the minor arc AB of the circumcircle of 4ABC such that PA, PB , PC are
all of integral length.
2419. Proposed by K.R.S. Sastry, Dodballapur, India.
Find all solutions to the alphametic:

M
D
E

I
B
S

X
A
U





E
S
M

1. The letters before the decimal points represent base ten digits, and addition is done in that base.
2. The letters after the decimal points represent base six digits, and addition is done in that base.
3. The same letter stands for the same digit, distinct letters stand for distinct digits, and initial digits are non-zero.
Readers familiar with cricket will realize that this is a real world problem!
Ed. Readers not familiar with cricket may be interested to learn that an
`over' consists of six `deliveries'!!
2420. Proposed by D.J. Smeenk, Zaltbommel, the Netherlands.
Suppose that x, y and z are integers. Solve the equation:

x2 + y2 = 2420z2 .

2421.

Proposed by Ice B. Risteski, Skopje, Macedonia.


What is the probability that the k numbers in the Las Vegas lottery on
a given payout day do not include two consecutive integers? The winning
numbers are an unordered random choice of k distinct integers from 1 to n,
where n k.

112

2422?.

Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.


Let A, B , C be the angles of an arbitrary triangle. Prove or disprove
that
p
1+1+1 
9 3
.
1=3

A B C

2 sin A sin B sin C 

2423. Proposed by Walther Janous, Ursulinengymnasium, Innsbruck, Austria.


Let x1 , x2 , : : : , xn 0 be real numbers such that x1 +x2 +: : :+xn = 1,
where n 2 is a natural number. Prove that
n
1 Y n , xk .
1+ x 
k
k=1 1 , xk

n
Y

k=1

Determine the cases of equality.


2424. Proposed by K.R.S. Sastry, Dodballapur, India.
In 4ABC , suppose that I is the incentre and BE is the bisector
of ABC , with E on AC . Suppose that P is on AB and Q on AC
such that PIQ is parallel to BC . Prove that BE = PQ if and only if
ABC = 2 ACB.
2425. Proposed by K.R.S. Sastry, Dodballapur, India.
Suppose that D is the foot of the altitude from vertex A of an acuteangled Heronian triangle ABC that is, one having integer sides and area.
Suppose that the greatest common divisor of the side lengths is 1. Find the
smallest possible value of the side length BC , given that BD , DC = 6.

NOTE OF THANKS
In the December 1998 issue of CRUX with MAYHEM 1998: 538
reference was made to whether a copy of the book Exercises de G om trie by
e e
F. Gabriel-Marie was available anywhere in Canada.
Dr. Kenneth Williams, Department of Mathematics and Statistics, Carleton University, has donated a copy to the Canadian Mathematical Society
for the use of the Editors of CRUX with MAYHEM. We are very grateful to
Dr. Williams for his generous gift. Having ready access to this book will be
of considerable assistance to the members of the Editorial Board of CRUX
with MAYHEM.

113

SOLUTIONS
No problem is ever permanently closed. The editor is always pleased to
consider for publication new solutions or new insights on past problems.
2255. 1997: 300; 1998: 378 Proposed by Toshio Seimiya, Kawasaki,
Japan.
Let P be an arbitrary interior point of an equilateral triangle ABC .
Prove that j PAB , PAC j  j PBC , PCB j.
The published solution 1998: 378-379 is incorrect. More precisely,
that solution deals with an inequality that is much simpler than what is given;
among other things, our given inequality does not extend to arbitrary isosceles triangles. We thank Toshio Seimiya for these remarks. Here is his solution.
II. Solution by Toshio Seimiya, Kawasaki, Japan.
Let M be the mid-point of BC . Then AM is the perpendicular bisector
of BC , and if P lies on it, the given relation holds with both sides zero.
We may therefore assume, without loss of generality, that P is an interior
point of 4ABM . We then have PAB
PAC and PBC PCB,
so that j PAB , PAC j = PAC , PAB and j PBC , PCB j =
PBC , PCB. Thus the relation we wish to prove reduces to
PAC , PAB
PBC , PCB .
1
Let Q be the re ection of P in the line AM ; then PAB = QAC and
PCB = QBC . Thus
PAC , PAB = PAC , QAC = PAQ ,
and
PBC , PCB = PBC , QBC = PBQ ,
so that 1 becomes
PAQ
PBQ .
2
Since PQ ? AM and AM ? BC we get PQkBC . Let PQ meet AB and
AC at R and S respectively, and let T be the re ection of B in RS. Then
PTQ = PBQ .
3
Since AM is the perpendicular bisector of both PQ and RS , the circumcentres of 4APQ and 4ARS lie on AM , so that the circumcircle of ABC
is tangent at A to the circumcircles of both 4ARS and 4APQ, which we
denote by , and ,0 respectively. Note that ,0 is contained in ,. Because
TRA = TRQ , ARQ = BRQ , ARQ = 120 , 60 = 60 , while
ASR = 60 , it follows that TRA = ASR. Hence RT is tangent to , so

114
that T is a point outside , and, consequently, T is a point outside ,0 . Since
A and T are on the same side of PQ, we have

PAQ

PTQ .

4

The desired relation 2 is a consequence of 3 and 4.


Summary of Seimiya's further comments.
The above argument extends to triangles for which B = C  60 ;
that is j PAB , PAC j  j PBC , PCB j also for isosceles triangles
when A  60 . On the other hand, when B = C 60 one can nd
positions for P where the given inequality fails to hold.

2309. 1998: 47 Proposed by Christopher J. Bradley, Clifton College, Bristol, UK.


Suppose that ABC is a triangle and that P is a point of the circumcircle,
distinct from A, B and C . Denote by SA the circle with centre A and radius
AP . De ne SB and SC similarly. Suppose that SA and SB intersect at P
and PC . De ne PB and PA similarly.
Prove that PA , PB and PC are collinear.
Solution by 12 of the 17 solvers; the notation of our featured solution
is by Florian Herzig, student, Cambridge, UK.
Since SA and SB are symmetric with respect to their line of centres
AB, their intersections P and PC are also symmetric with respect to this
line. Let QC be the mid-point of PPC , etc. By the above conclusion, QC is
the foot of the perpendicular from P onto AB . The points QA , QB and QC
are collinear since they determine the Simson or Wallace line of P with
respect to 4ABC . A dilatation with centre P and factor 2 takes QA to PA ,
etc. Hence PA , PB and PC are collinear as well in a line parallel to the
Simson line .
Also solved by MICHEL BATAILLE, Rouen, France; NIELS BEJLEGAARD, Stavanger,
Norway; FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid, Spain; MANSUR
BOASE, student, Cambridge, England; GORAN CONAR, student, Gymnasium Vara din,
z
Vara din, Croatia; NIKOLAOS DERGIADES, Thessaloniki, Greece; WALTHER JANOUS, Ursuz
linengymnasium, Innsbruck, Austria; MICHAEL LAMBROU, University of Crete, Crete, Greece;
GERRY LEVERSHA, St. Paul's School, London, England; JOEL SCHLOSBERG, student, Bayside, NY, USA; TOSHIO SEIMIYA, Kawasaki, Japan; D.J. SMEENK, Zaltbommel, the Netherlands; PARAYIOU THEKLITOS, Limassol, Cyprus; G. TSINTSIFAS, Thessaloniki, Greece; JOHN
VLACHAKIS, Athens, Greece; and the proposer.
Schlosberg found the problem in a slightly di erent form on page 43 of Ross
Honsberger's Episodes in 19th and 20th Century Euclidean Geometry. Bellot informs us that
the line PA PB PC is called the Steiner line of P with respect to 4ABC ; its properties are discussed in Y. and R. Sortais, La g om trie du triangle, Exercises r solus, Hermann, Paris, 1987.
e e
e
Both he and Seimiya report that the line passes through the orthocentre of 4ABC .

115

2312.

1998: 47 Proposed by K.R.S. Sastry, Dodballapur, India.


r
The r n gonal number is given by P n;r = n , 2 r2 , n , 4 2 ,
where n  3, r = 1, 2, : : : .
Prove that, in the interval P n;r; P n; r + 1 , there is an n , 1
gonal number.
Solution by Florian Herzig, student, Cambridge, UK.
Assume that for some n  3 and r, s  1,
th

P n , 1; s P n;r P n;r + 1 P n , 1; s + 1 .
Therefore P n , 1; s + 1 , P n , 1; s P n;r + 1 , P n;r.

1

Now


2
2
P n;r + 1 , P n;r = n , 2 r + 1 , r , n , 4 1
2
2
= rn , 2 + 1 .
Similarly, P n , 1; s + 1 , P n , 1; s = sn , 3 + 1.
Hence, from the last inequality above, sn , 3+1 rn , 2+1, and thus
n,
n  4 and s rn , 32 .
Using the fact that P n , 1; s, as a real function in s, is strictly increasing for all s  1 veri ed by di erentiation, we have

2 2
n,
2
P n , 1; rn , 32 = n , 2 r2 , n ,n ,n , 5 r
n , 3
 3 2
 P n;r;
n , 22  n , 2 and n , 2n , 5  n , 4 .This contradicts 1
since
n,3
n,3
and hence it follows that in P n;r; P n;r + 1 there is an n , 1 gonal

P n , 1; s

number.

Also solved by MANSUR BOASE, student, Cambridge, England;


CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; RICHARD I. HESS, Rancho
Palos Verdes, California, USA; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria;
MICHAEL LAMBROU, University of Crete, Crete, Greece; and the proposer.
It is worth noting that all seven solutions involved di erent methods. Boase showed

lq
m
that P n , 1; n,2 always lies in the interval P n; r; P n; r +1 . Janous and others
n,3
noted that the involvement of n , 1 gonal numbers implies that n , 1  3; that is, n  4.
Referring to the geometry involved, the proposer noted that the problem can be reformulated:
Prove that there are just one or just two n , 1 gonal numbers in P n; r; P n; r + 1 .

116

2313. 1998: 47 Proposed by Heinz-Jurgen Sei ert, Berlin, Germany.


Let N be a non-negative integer and let a and b be complex numbers
with a, b 62 f0, ,1, ,2, : : : , ,n , 1g. Find a closed form expression for
n
X

,1k ,
k=0 ak bn,k

where ak denotes the Pochhammer symbol, de ned by a0 = 1,


ak = aa + 1 : : : a + k , 1, k 2 N.
Solution by Michael Lambrou, University of Crete, Crete, Greece.
If S a; b; n denotes the sought sum, we show by induction on n that

, n
+1
Sa; b;n = a ,a1nb ++ , 1 ab ,b1n+1 .
 + n 2 n n

Indeed, for n = 1 it is true since

1
+b,
b
Sa; b; 1 = 1 , a = a, bba, 1ab 1 = a ,a1ab, 1, 1b .
b
a+
 + , ab

The induction step uses the facts

cc + 1t = ct+1 and c + tct = ct+1


for all c and t, and the observation
Sa; b; n + 1
n+1
n
X
X
,1k
,1k
, n+1
=
= 1 a b + 1 + a1
b k=0 k
n+1
n,k
k=0 ak bn+1,k
, n+1
= 1 Sa; b + 1; n + a1
b
n+1
n  +1 , n+1
, +1
= 1 a+ b1nn ,+ ,1 b+n1 + a1
b a + 1an b n
n+1
nbn+1a + n + ,1n+1a + b + n , 1bn+1
= a , 1n+2 + ,1 a + b + n , 1a b
n+1 n+1
a , 1n+2 + ,1n+1b , 1n+2 ,
= a + b + n , 1a b
n+1 n+1
as required.

Also solved by KEE-WAI LAU, Hong Kong; and the proposer. One other reader submitted
a solution which was not in closed form.
Lau notes that a + b + n , 2 6= 0 is required for this problem, and that if a + b + n , 2 = 0
then the given expression likely does not have a closed form.

117
n
X
The proposer gives a similar expression for ,1kak bn,k , and lists some comk=0
binatorial identities as special cases of these two results. For instance, from the formula for
S 1=2; 1=2; 2n and the observation that 1=2k = 4,k 2k!=k! he obtains
, 
2n
X ,1k 4n
1
;
, 2n2k =
1 , 2n
k=0
k
n
X
and as a special case of his formula for ,1kak bn,k he gets
k=0
2n
X ,1k
2n + 1
,2n =
n+1 .
k=0 k

2314.

1998: 107 Proposed by Toshio Seimiya, Kawasaki, Japan.


Given triangle ABC with AB AC . The bisectors of angles B and C
meet AC and AB at D and E respectively, and DE intersects BC at F .
1
Suppose that DFC = 2  DBC , ECB . Determine angle A.
Solution by Florian Herzig, student, Cambridge, UK.
First we show that DI = IE where I is the incentre. We have

1 DIC = 1  DBC + ECB  = DFC + ECB


2
2
= ECF + EFC = DEC .
Since DIC = DEC + EDB , it follows that DI = EI . Now by the
Sine Law:

AI = ID = IE = AI
sin ADI
sin AEI
sin A
sin A
2
2
and hence either ADI = AEI or ADI + AEI = 180 . From the rst
of these it easily follows that the angles B and C are equal, a contradiction.
From the second it follows that the quadrilateral AEID is cyclic. Hence


180 = A + EID = A + 90 + A ,

and hence A = 60 .

Also solved by FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid, Spain;
MANSUR BOASE, student, Cambridge, England; CHRISTOPHER J. BRADLEY, Clifton College,
Bristol, UK; GORAN CONAR, student, Gymnasium Vara din, Vara din, Croatia; NIKOLAOS
z
z
DERGIADES, Thessaloniki, Greece; DAVID DOSTER, Choate Rosemary Hall, Wallingford, Connecticut, USA; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; KEE-WAI LAU,
Hong Kong; GERRY LEVERSHA, St. Paul's School, London, England; D.J. SMEENK, Zaltbommel, the Netherlands; PARAYIOU THEOKLITOS, Limassol, Cyprus; and the proposer. There
were two incorrect solutions.

118

2315. 1998: 107 Proposed by V aclav Konecn , Ferris State Uniy


versity, Big Rapids, Michigan, USA. s
1 n,1 , where F n is the

Prove or disprove that F n = n 1 ,


n
maximum value of

f x1; x2; : : : ; xn = sin x1 cos x2 : : : cos xn + cos x1 sin x2 : : : cos xn


+ : : : + cos x1 cos x2 : : : sin xn ,
xk 2 0; =2 , k = 1, 2, : : : , n, and n 1 is a natural number.
Solution by Heinz-Jurgen Sei ert, Berlin, Germany.
Let n  2 and let

n
C n = n,p,n1,2 .
n
Then 1998: 119 if a1 , : : : , an  0,
n
n
Xp
Yp
ak 
ak + C n .
1

k=1

k=1

Replacing ak by C nbk and noting that C nn,1=2 = F n give


n
Xp

k=1

bk  F n

n
Y p

k=1

bk + 1 ,

1

valid for all b1, : : : , bn  0.


p
If x1 , : : : , xn 2 0; =2, let bk = Qn 2 xk . Since bk + 1 = sec xk ,
tan
we obtain, from 1, after multiplying by j =1 cos xj ,
n
X

k=1

sin xk

n
Y

j =1;j 6=k

cos xj  F n .

Clearly, the latter inequality remains valid if xk = =2 for some k. Since
there is equality if xk = arcsin1=pn for all k, it follows that F n, in fact,
is the maximum value of the considered expression.

Also solved by CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; THEODORE


CHRONIS, student, Aristotle University of Thessaloniki, Greece; NIKOLAOS DERGIADES, Thessaloniki, Greece; RICHARD I. HESS, Rancho Palos Verdes, California, USA; WALTHER JANOUS,
Ursulinengymnasium, Innsbruck, Austria; MICHAEL LAMBROU, University of Crete, Crete,
Greece; and PANOS E. TSAOUSSOGLOU, Athens, Greece.
The problem was inspired by problem 2214. Janous's solution also made use of 2214.

119

2316.

1998: 108 Proposed by Toshio Seimiya, Kawasaki, Japan.


Given triangle ABC with angles B and C satisfying C = 90 + 1 B .
2
Suppose that M is the mid-point of BC , and that the circle with centre A
and radius AM meets BC again at D. Prove that MD = AB .
Solution by David Doster, Choate Rosemary Hall, Wallingford, Connecticut, USA.
Let E be the other point where the circle centred at A with radius AC
meets line BC . Then AEC = ACE , so that AED = ACM: Also,
AD = AM and ADE = AMC . Therefore, 4ADE  4AMC . Hence,
=
DE = CM . But CM = MB; therefore, DE = MB, so that MD = BE.
Note that AEC = ACE = 90 , 1 B . Hence,
2

1 B = 90 , 1 B .
BAE = 180 , B + 90 , 2
2
Thus, AB = BE . Therefore, MD = AB .

Also solved by MIGUEL AMENGUAL COVAS, Cala Figuera, Mallorca, Spain; FRANCISCO
BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid, Spain; MANSUR BOASE, student, Cambridge,
England; CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; GORAN CONAR, student, Gymnasium Vara din, Vara din, Croatia; NIKOLAOS DERGIADES, Thessaloniki, Greece;
z
z
FLORIAN HERZIG, student, Cambridge, UK; WALTHER JANOUS, Ursulinengymnasium, Inns

bruck, Austria; VACLAV KONECNY, Ferris State University, Big Rapids, Michigan, USA;
MICHAEL LAMBROU, University of Crete, Crete, Greece; GERRY LEVERSHA, St. Paul's School,
London, England; GOTTFRIED PERZ, Pestalozzigymnasium, Graz, Austria; JOEL SCHLOSBERG,
student, Robert Louis Stevenson School, New York, NY, USA; RON SHEPLER, Ferris State University, Big Rapids, Michigan, USA; D.J. SMEENK, Zaltbommel, the Netherlands; PARAYIOU
THEOKLITOS, Limassol, Cyprus; PANOS E. TSAOUSSOGLOU, Athens, Greece; PAUL YIU,
Florida Atlantic University, Boca Raton, Florida, USA; and the proposer.

2317. 1998: 108 Proposed by Richard I. Hess, Rancho PalosVerdes,


California, USA.

2
3

c
e

The quadrilateral shown at the left


has integer elements a through e.
The angles as shown are integer
multiples of the smallest.
a What is the smallest possible
value of c?
b What is the smallest possible
value of c if must be obtuse?

Solution by Michael Lambrou, University of Crete, Crete, Greece.


We show that the answer to a is c = 5  17  29  1049 = 2585785 and
to b it is c = 19  41  59  139  50539 = 322872394081.

120
a From the Law of Sines on the upper triangle we have

a = b = c
sin
sin 2
sin 3
and as sin2 = 2 sin cos and sin3 = 4 cos2 , 1 sin , this reduces
to
b
a = 2 cos = 4 cos2c , 1 .
Eliminating cos we nd cos = b=2a and so
a2 + ac , b2 = 0 .
1
As a quadratic in a this last must have discriminant a perfect square, say x2 .
That is,

c2 + 2b2 = x2 .
2
Hence there exist s, t, m 2 N with s; t = 1 and
c = s2 , t2m , 2b = 2stm , x = s2 + t2m .
3
Note that if c is odd this is certainly so, but when c is even we may have the
dual relations:

2b = s2 , t2m ,

c = 2stm ,

x = s2 + t2 m .
However, this amounts to the same thing because then, by 2, x is even; so
either m is even or s2 + t2 is even. In the rst case set m = 2M , s1 = s + t,
t1 = s , t and we then have c = s2 , t2M , b =2s1t12M , x = s2 + t2M
1
1
1
1
which is the same as 3. If, instead, m is odd, so s + t is even, we see that
s, t are of the same parity; set s1 = s + t=2, t1 = s , t=2, M = 2m,
and we recapture 3, since s1 , t1 , M 2 N.
From 1 keeping the positive sign only we have

,c + px2 = t2m , so cos = b = s .


a=
2
2a 2t
From the lower triangle we have 0 3 + 4 180 ; that is
1 cos cos   0:90096887 0:9 .
7
Using cos = s=2t we have
1:8t 2t cos  s 2t .
4
7
In particular t cannot be 1, 2, 3, 4, or 5, as there is no integer s in the open
,

interval 2t cos  ; 2t . The least allowable t is t = 6 whence s = 11. For
7
the record the next few allowable pairs are t; s = 7; 13, 8; 15, 9; 17.

121
From the Law of Sines on the lower triangle we have

e
c
= sin4 = sin7 .
sin 3
Hence d sin 4 = e sin 3 or
d8 cos3 , 4 cos  sin = e4 cos2 , 1 sin .
Using cos = s=2t we get
dss2 , 2t2 = ets2 , t2 .
5
Recall that s; t = 1. We show that we also have
s; s2 , t2 = t; s2 , 2t2  = s2 , 2t2 ; s2 , t2  = 1 .
Indeed, if p is a prime and pjs, pjs2 , t2 , then pjt2 , so pjt. Thus pjs; t = 1,
showing that s; s2 , t2  = 1. Similarly t; s2 , 2t2  = 1, and if pjs2 , 2t2 ,
pjs2 , t2, then pjs2 , t2 , s2 , 2t2 = t2, so pjt, etc., as before.
These conditions of relative primality applied to 5 show that there
exists a constant  2 N such that

d = ts2 , t2

Now

and

e = ss2 , 2t2 .

6

6
80 cos4 24 cos2
sin
c = dsin37 = d64 cos ,4 cos2 ,+1 sin , 1 sin

6
4 2 +
2 4
6
= s , 5s t t3 6s t , t  by 6 .

From 3 we have

mt3s2 , t2 = s6 , 5s4t2 + 6s2t4 , t6 .

7

Observe that

t; s6 , 5s4t2 + 6s2t4 , t6 = 1 = s2 , t2; s6 , 5s4t2 + 6s2 t4 , t6  .


Indeed the rst, as above, is clear. For the second, note that

s6 , 5s4t2 + 6s2t4 , t6 = s2 , t2s4 , 4s2t2 + 2t4 + t6 ,


so any prime dividing both terms must divide s2 , t2 and t6 and so t, which
is impossible, as above. We conclude from 7 that there is a 2 N such that
m = s6 , 5s4t2 + 6s2t4 , t6  ,  = t3s2 , t2 ,

122
and hence

c = s2 , t2m = s2 , t2s6 , 5s4t2 + 6s2t4 , t6


= s2 , t2s3 + s2t , 2st2 , t3 s3 , s2t , 2st2 + t3  .
For the least c, we clearly need = 1 any solution for s; t;  is larger than
for s; t; 1. We shall show that the least c comes from t = 6, s = 11 and
= 1.

Denote by ct; s the value of c at the pair t; s and = 1. For
6  t  12 the relatively prime pairs t; s satisfying 4 are 6; 11, 7; 13,
8; 15, 9; 17, 10; 19, 11; 20, 11; 21, and 12; 23. They give
c6; 11 = 2585785 the least, c7; 13  1:7  107, c8; 15  7:4  107,
c9; 17  2:4  1089, c10; 19  6:6  108, c11; 20  1:6  108,
c11; 21  1:6  10 , and c12; 23  3:6  109. We now show that
for t  13 we still get larger values than c6; 11 by eliminating c as follows.
By 4 we have for the factors of c:

s2 , t2  1:8t2 , t2 = 2:24t2 ;
s + s t , 2st2 , t3  1:8t3 + 1:8t2 , 22tt2 , t3
= 4:072t3 .
The other factor must be positive since c, and thus the product of the re3

maining two factors, is positive. As this factor is an integer, it must be at


least 1. Hence
c  2:24t24:072t31 9t5 .
Hence if t  13 we have c 9  135  3:3  106 c6; 11.
To conclude part a the least value of c is c6; 11 = 2585785.
b If we insist that is obtuse, we have equivalently 3 + 4 90 ;

that is, 14 . So instead of 4 we have the sharper requirements:

s = cos cos   0:9749279 ;


14
2t

that is,

1:949t


2t cos 14

2t .

8

It follows that t  20, as for 1  t  19 there is no s 2 N in the open


interval 1:949t; 2t. The least t is t = 20 with corresponding s = 39 and
the next few pairs are 21; 41, 22; 43, 23; 45, : : : . We show that the
least value of c is c20; 39 = 19  41  59  139  50539  3:2  1011 .
Certainly, c20; 39 c21; 41 as the expression for c gives c21; 41 =
1240  58799  6719  4:89  1011. Moreover, for larger t we use estimates

123
arising from 8, namely

s2 , t2
s3 + s2t , 2st2 , t3

1:949t2 , t2 2:798t2 ;
1:949t3 + 1:949t2t , 22tt2 , t3
6:202t3 ;
s3 , s2t , 2st2 + t3
1:949t3 , 2t2t , 22tt2 + t3
0:4034t3 .
Multiplying together we nd c
7t8 . Thus for t  22 we have
8  3:8  1011
c 7  22
c20; 39. This concludes the proof.
Also solved by CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; KENNETH
M. WILKE, Topeka, Kansas, USA; and the proposer.

2318. 1998: 108 Proposed by V aclav Konecn y, Ferris State University, Big Rapids, Michigan, USA.
Suppose that ABC is a triangle with circumcentre O and circumradius R.
Consider the bisector ` of any side say AC , and let P the pedal
point" be any point on ` inside the circumcircle.
Let K , L, M denote the feet of the perpendiculars from P to the lines
AB, BC , CA respectively.
Show that KLM the area of the pedal triangle KLM  is a decreasing
function of = OP , 2 0; R.
Combination of the solutions by Francisco Bellot Rosado, I.B. Emilio
Ferrari, Valladolid, Spain; Michael Lambrou, University of Crete, Crete,
Greece; Gerry Leversha, St. Paul's School, London, England; and Toshio
Seimiya, Kawasaki, Japan.
Either by referring to problem 2236, or R.A. Johnson's Advanced Euclidean Geometry, Dover, 1960, theorem 198, page 139, it is known that

2
, 2
KLM = R 4ROP ABC .
2
Thus, it follows that KLM is a decreasing function of .

Also solved in full by CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; GORAN
CONAR, student, Gymnasium Vara din, Vara din, Croatia; NIKOLAOS DERGIADES, Thessaz
z
loniki, Greece; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; D.J. SMEENK,
Zaltbommel, the Netherlands; PARAYIOU THEOKLITOS, Limassol, Cyprus; and the proposer.
Bellot Rosado and Leversha note that there is no need for P to lie on a perpendicular
bisector of a side.
Dergiades and Janous observe that if d = R, then KLM = 0, and so K , L, M lie on
the Simson line.

124
Dergiades recalls the result as being known as Steggals' Theorem. He observes that if
= E and if E0 , E1 , E2 and E3 are the areas of the triangle KLM when P is the
incentre, and the three excentres respectively, then we can deduce, from Euler's Theorem, that
d2 = RR , 2r d2 = RR , 2r1  d2 = RR , 2r2  d2 = RR , 2r3  .
1
2
3
Steggal's Theorem now gives
E0 = r , E1 = r1 , E2 = r2 , E3 = r3 ,
E
2R
E
2R
E
2R
E
2R
and since r1 + r2 + r3 = r + 4R Bobillier's Theorem, we have
E1 + E2 + E3 = E0 + 2E .

ABC

UK.

2319.

1998: 108 Proposed by Florian Herzig, student, Cambridge,

Suppose that UV is a diameter of a semicircle, and that P , Q are two


points on the semicircle with UP UQ. The tangents to the semicircle at P
and Q meet at R. Suppose that S is the point of intersection of UP and V Q.
Prove that RS is perpendicular to UV .
Editorial note. Some solvers sent in more than one correct solution.
They are indicated by a y after their names.
I. Solution by Diane and Roy Dowling, University of Manitoba,
Winnipeg, Manitoba.
The problem may be generalized slightly as follows: UV is a diameter
of a circle; P and Q are distinct points on the circle; PQ is not a diameter;
P 6= U and Q 6= V ; the tangents to the circle at P and Q meet at R; the lines
UP and V Q meet at S; prove that RS is perpendicular to UV . The given
conditions ensure that the points of intersection R and S exist, are unique
and distinct from each other.
Choose coordinate axes and scale so that the origin O is the centre of
the circle, V = 1; 0 and consequently U = ,1; 0. Let P = a; b; then
a2 + b2 = 1. Let Q = c; d; then c2 + d2 = 1. Let R = x1; y1 and
S = x2; y2.
The equation of PR is ax + by = 1 and the equation of QR is
cx + dy = 1. Since PR and QR intersect at exactly one point, the determinant of the coe cients of this system is non-zero. Solving the system
for x we get the x coordinate of R:

d
x1 = ad , b .
, bc
The equation of UP is bx , a + 1y = ,b and the equation of V Q
is dx , c , 1y = d. Since UP and V Q intersect at exactly one point,
namely S , the determinant of the coe cients of this system is also non-zero.
Solving the system for x we get the x coordinate of S :
bc
x2 = ad + bc + d , b .
ad , + d + b

125
Therefore

2
d
+
d2 , a2  2
x1 , x2 = ad , b , ad , bc + d , b = ad 1 bcad, bbc1 , c+b
, bc ad bc + d + b
,
, +d
2 2
2 2
d
d
= ad , bcbad, bbc + d + b = 0 .
,
It follows that RS is perpendicular to UV .
II. Solution by Toshio Seimiya y, Kawasaki, Japan.
Let T be the intersection of UQ and V P , and let M be the mid-point
of ST . Since UPV = UQV = 90 , T is the orthocentre of triangle SUV ,
so that ST ? UV . Hence we have PTS = PUV . Since SPT = 90
and M is the mid-point of ST we get MP = MT . Thus
MPV = MPT = MTP = PTS = PUV .
Hence PM is tangent to the semicircle. Similarly QM is tangent to the semicircle. Therefore M coincides with R. As ST ? UV , we have RS ? UV .
III. Solution by Keivan Mallahi y, student, Sharif University of Technology, Tehran, Iran.
Let T be the intersection of UQ and PV . Consider the hexagon
QQUPPV . Note that all of its vertices lie on the semicircle. Applying Pas-

cal's Hexagon Theorem we see that the following points are collinear:

QQ PP = R , QU PV = T , UP V Q = S
where QQ and PP are intended to be the tangents at Q and P , respectively. Thus it is su cient to show that the line ST is perpendicular to UV .
To this end note that UQV = V PU = 90 , so the lines UQ and PV
are the altitudes of the triangle SUV . Since the altitudes of a triangle are
concurrent, ST must be the third altitude, which completes the proof.
IV. Solution by Michael Lambrou y, University of Crete, Crete, Greece.
Consider the nine-point Feuerbach circle of triangle UV S . This passes
through P and Q as P and Q are the feet of the perpendiculars from V
and U , respectively, because UPV = UQV = 90 , being angles on the
semicircle. If PV , QU meet at T , the orthocentre, then ST is also an altitude meeting UV at D, say. Note that the nine-point circle also passes
through D, through the mid-point M of ST , and through the mid-point O
of UV so O is the centre of the semicircle on UV . Moreover, MO is a
diameter of the nine-point circle as MDO = 90 . Thus we also have
PM ? PO. But OP is a radius of the semicircle on UV , so PM , being perpendicular to PO at its endpoint P , is a tangent to the semicircle. Similarly
QM is a tangent. In other words the point M is where these two tangents
meet. Thus M and R coincide and clearly SR being the same as SM  is an
altitude of SUV . This conclude the proof that SR ? UV .

126
Also solved by MIGUEL AMENGUAL COVAS, Cala Figuera, Mallorca, Spain; MICHEL
BATAILLE, Rouen, France; FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid,
Spain; MANSUR BOASE, student, Cambridge, England; CHRISTOPHER J. BRADLEY, Clifton
College, Bristol, UK; GORAN CONAR, student, Gymnasium Vara din, Vara din, Croatia;
z
z
NIKOLAOS DERGIADES, Thessaloniki, Greece; DAVID DOSTER, Choate Rosemary Hall, Wallingford, Connecticut, USA; STERGIOU HARAFAPOS y, Greece; RICHARD I. HESS, Rancho
Palos Verdes, California, USA; WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; MASOUD KAMGARPOUR, Carson Graham Secondary School, North Vancouver, British

Columbia; KATHLEEN E. LEWIS, SUNY Oswego, Oswego, NY, USA; JOSE H. NIETO, Universidad del Zulia, Maracaibo, Venezuela; GOTTFRIED PERZ, Pestalozzigymnasium, Graz, Austria; D.J. SMEENK, Zaltbommel, the Netherlands; PARAYIOU THEOKLITOS, Limassol, Cyprus;
GEORGE TSAPAKIDIS, Agrinio, Greece; PAUL YIU, Florida Atlantic University, Boca Raton,
Florida, USA; and the proposer y. There was one incorrect solution submitted.
Although four methods of solving are given above, this was not exhaustive. Bataille used
inversion and Theoklitos used radical axes in their proofs.

2320. 1998: 108 Proposed by D.J. Smeenk, Zaltbommel, the Netherlands.


Two circles on the same side of the line ` are tangent to it at D. The
tangents to the smaller circle from a variable point A on the larger circle
intersect ` at B and C . If b and c are the radii of the incircles of triangles
ABD and ACD, prove that b + c is independent of the choice of A.
Solution by Florian Herzig, student, Cambridge, UK.
Let m be the parallel to ` which is also tangent to the smaller circle.
We will show that
b + c is constant if A is a point outside the region between `
and m so that D is between B and C , and b : c is constant if A
is between these lines.
Denote the larger circle by C1M ; R and the smaller circle by C2 N ; r.
Let S 2 AB and T 2 AC be the points of contact with C2 . Also let IB
and IC be the incentres of triangles ABD and ACD. U and V are the feet
of the perpendiculars from IB and IC to `. We will now calculate the ratio
AS : AD which is constant in both cases: let P be the second intersection of
C2 and AD. Then AP  AD = AS2. Also by similarity PD : AD = r : R.
Hence


AS2 = AP  AD = AD2 1 , r ,
R

and thus the ratio AS : AD is constant for all positions of A.


If A is not between ` and m, then

DU = BD + AD , AB = AD , AS = CD + AD , AC = DV
2
2
2
because BD = BS , CD = CT and AS = AT . Therefore the two
incircles touch line AD at the same point, call it Y . It also follows that

127

IB IC ? AD. Moreover DN is the middleTparallel between IB U and IC V ,


whence 2DX = b + c, where X = DN IB IC . If L is the mid-point of
AD, then 4DLM  4DY X so that DX : DM = DY : DL.
Using 2DX = b + c, DM = R, and 2DY = 2DU = AD , AS it
follows that

2RAD , AS  = 2R 1 , r1 , r = 2R , pR2 , Rr 


b+c =
AD
R

is constant.
Otherwise, if A is between lines ` and m assume, without loss of
generality, that DB DC . Then 2DV = AD , AS as before, but

DU = AD + BD , AB = AD + AS .
2
2
pR2 , Rr
AD + AS = R + p
Therefore b : c = DU : DV =
AD , AS R , R2 , Rr is constant.

Finally, note that all the calculations of the rst case remain valid
mutatis mutandis if instead of the incircle in the second case we take the
excircle opposite B so that b is the radius of the excircle of 4ABD that
touches the side AD .

Also solved by FRANCISCO BELLOT ROSADO, I.B. Emilio Ferrari, Valladolid, Spain;
CHRISTOPHER J. BRADLEY, Clifton College, Bristol, UK; GORAN CONAR, student, Gymnasium Vara din, Vara din, Croatia; NIKOLAOS DERGIADES, Thessaloniki, Greece; WALTHER
z
z
JANOUS, Ursulinengymnasium, Innsbruck, Austria; MICHAEL LAMBROU, University of Crete,
Crete, Greece; GERRY LEVERSHA, St. Paul's School, London, England; TOSHIO SEIMIYA, Kawasaki, Japan; PARAYIOU THEOKLITOS, Limassol, Cyprus; HOE TECK WEE, Singapore; and
the proposer.
Bellot found the problem in the Japanese Encyclopedia of Geometry, volume 4, page 261,
e
which provided the reference Journal de Math matiques El mentaires 34 1909, problem 7000.
e
Several readers noted that the statement of our problem was not entirely correct, but Seimiya
was the only solver besides Herzig to provide a correct alternative using an excircle.

2323.

1998: 109 Proposed by K.R.S. Sastry, Dodballapur, India.


Determine a positive constant c so that the Diophantine equation

uv2 , v2 , uv , u = c
has exactly four solutions in positive integers u and v .

Solution by Digby Smith, Mount Royal College, Calgary, Alberta.


Since u = ,1 , c 0 when v = 1, we have v  2. Then

v2 , v , 1 = v , 22 + 3v , 2 + 1

128
and

2
c
u = v2v, + , 1
v

1

When c = 61, one veri es easily that 1 admits no positive integer
solutions for u if v = 5, 6, 7, 8, while v = 2, 3, 4, 9 would yield four
solutions in u and v : u; v  = 65; 2, 14; 3, 7; 4 and 2; 9. Finally, if
v 9, then v , 9v + 7 0 would imply that 2v2 , v , 1 v2 + 61,
+61
making u = vv ,v,1 2. However, if u = 1, then v = ,62 0. Hence 1
admits no positive integer solutions for u if v 9. Therefore 1 has exactly
four solutions as listed above.
2

Also solved by SAM BAETHGE, Nordheim, Texas, USA; MANSUR BOASE, student,
St. Paul's School, London, England; CHRISTOPHER J. BRADLEY, Clifton College, Bristol,
UK; THEODORE CHRONIS, student, Aristotle University of Thessaloniki, Greece; FLORIAN
HERZIG, student, Cambridge, UK; RICHARD I. HESS, Rancho Palos Verdes, California, USA;
WALTHER JANOUS, Ursulinengymnasium, Innsbruck, Austria; MICHAEL LAMBROU, University of Crete, Crete, Greece; KEE-WAI LAU, Hong Kong; GERRY LEVERSHA, St. Paul's
School, London, England; DAVID E. MANES, SUNY at Oneonta, Oneonta, NY, USA; PANOS E.
TSAOUSSOGLOU, Athens, Greece; and the proposer. There was also one incorrect solution.
While most solutions gave only one value of c, Manes gave two. Baethge and Chronis
each gave three values of c, while Hess gave the following twelve values and claimed that they
are the twelve lowest" values of c: 51, 61, 156, 321, 336, 402, 431, 486, 526, 611, 761 and
771. In terms of the frequency" of the value of c given, the top three are c = 61, 336 each
given six times and c = 51  given ve times. The largest value of c given was c = 26461,
obtained by Janous. Both Hess and Herzig considered similar problems in which one seeks
the value of c for which the given equation has exactly n solutions for n = 4, 5, 6, : : : .
Herzig used a computer to nd that, for 4  n  10, the corresponding minimum values of c
are 51, 1381, 3966, 33776, 51816 and 14686766, respectively, while Hess list the ten lowest
values of c for each n such that 4  n  8. His lowest values agree with Herzig's.
For those curious, the editor has obtained by computer the next few lowest values of c
for n = 4: 776, 816, 1066, 1071, 1146, 1153, 1172, 1201, 1271 and 1360 as those less than
1381, the lowest value with ve solutions.

Crux Mathematicorum

Founding Editors R dacteurs-fondateurs: L opold Sauv & Frederick G.B. Maskell


e
e
e
Editors emeriti R dacteur-emeriti: G.W. Sands, R.E. Woodrow, Bruce L.R. Shawyer
e

Mathematical Mayhem

Founding Editors R dacteurs-fondateurs: Patrick Surry & Ravi Vakil


e
Editors emeriti R dacteurs-emeriti: Philip Jong, Je Higham,
e
J.P. Grossman, Andre Chang, Naoki Sato, Cyrus Hsia